planar hall sensor for influenza immunoassay · s=nc = 2 is found for the magnetic immunoassay, and...

General rights Copyright and moral rights for the publications made accessible in the public portal are retained by the authors and/or other copyright owners and it is a condition of accessing publications that users recognise and abide by the legal requirements associated with these rights.

Users may download and print one copy of any publication from the public portal for the purpose of private study or research.

You may not further distribute the material or use it for any profit-making activity or commercial gain

You may freely distribute the URL identifying the publication in the public portal If you believe that this document breaches copyright please contact us providing details, and we will remove access to the work immediately and investigate your claim.

Downloaded from orbit.dtu.dk on: Apr 28, 2021

Planar Hall Sensor for Influenza Immunoassay

Ejsing, Louise Wellendorph

Publication date:2006

Document VersionPublisher's PDF, also known as Version of record

Link back to DTU Orbit

Citation (APA):Ejsing, L. W. (2006). Planar Hall Sensor for Influenza Immunoassay. Technical University of Denmark.

https://orbit.dtu.dk/en/publications/631f32df-cd16-4764-9a50-b0605f3c6391

Ph.D. Thesis, s021568

Planar Hall sensor for influenzaimmunoassay

Louise Wellendorph Ejsing

Supervisors: Aric K. Menon and Mikkel F. Hansen

MIC – Department of Micro and NanotechnologyTechnical University of Denmark

May 5, 2006

Abstract

The title of this thesis is Planar Hall sensor for influenza immunoassay. The thesis con-siders fabrication, characterization and demonstration of planar Hall sensors for influenzaimmunoassay detection.

The goal of this research project is, first of all, to design a magnetic sensor capable ofdetecting magnetic beads. These beads are polystyrene spheres with sizes ranging betweena few hundred nanometers to a few micrometers, and can be magnetized in an applied field.In order to integrate this sensor into a device for clinical tests the sensing principle hasto be sensitive as well as reliable. A signal-to-noise study for various sensor types, GMR,spin-valve, and planar Hall sensors, points towards the planar Hall effect as a promisingsensing principle for DC detection. DC detection will probably be easier to handle on achip than AC detection.

A second goal is to demonstrate a relevant application, where conventional techniqueshave failed. Statens Serum Institut has provided antibodies and antigens for the demon-stration of influenza detection.

The theoretical analysis of single bead signal shows that the planar Hall sensor haspotential for single bead detection. The active area of the sensor has to be designed tomatch the specific bead, and using areas below 1µm×1µm offers the possibility of detectinga single 50 nm bead. The theoretical detection limit as a function of sensor size presentsgreat possibilities for the planar Hall sensor, the main reason being its low noise level.

Following the single bead study, theoretical investigation of the specific sensor andbead combination used in the experimental part of this thesis is presented. The fabricatedsensors 20µm×20µm are used with 250 nm Nanomag-D beads. These sensors and beads areused for the influenza experiments. The approximate theoretical signal from a monolayerof beads is β ≈ 0.024(ξoutside−ξsensor), where ξ is the layer coverage. β is the field producedby the beads normalized to the applied field and can be compared to the experimentaldata.

Furthermore, the effect of screening the positive contribution from the total signal byplacing a simple barrier adjacent to the sensor cross is evaluated. The barrier can beconstructed in SU-8, which is used for attachment of biochemical species. Theoretically,the signal from a monolayer of magnetic beads can be enhanced by this procedure. Ad-ditionally, the capture of beads by fringing fields produced at the voltage leads can bereduced. The fringing fields capture beads at areas insignificant to the measurements butbiologically active material would be lost at these capture sites.

Next, design, fabrication and characterization of planar Hall sensors for the influenza

iii

iv ABSTRACT

immunoassay is presented. The chip design match the prospect of detecting just a few 250nm beads, and is prepared for use in biological experiments.

First the fabrication and results with nickel sensors are presented. Based on theseresults, the work with exchange bias for constructing an easy direction instead of aneasy axis in permalloy planar Hall sensors continues. The nickel sensors are fabricated inMIC’s clean room and constitute a compromise with respect to magnetic material, sinceonly nickel is available in the clean room. However, the nickel sensors prove the conceptof magnetic bead detection with planar Hall sensors.

The permalloy sensors are optimized with respect to the magnetic material and designof an easy direction. In an easy axis two equivalent magnetic states are present, whichgive identical electric signals with opposite signs. Exchange coupling of the easy axis toan antiferromagnet yields an easy direction. Thus the magnetization has always the samestarting point in the absence of an applied field. This principle is utilized in the optimizedsensors. Using DC sensor currents, these sensors can measure magnetic beads without afield applied. The magnetic field generated by the current is sufficient to magnetize thebeads, which will be an advantage when designing a point-of-care chip.

Using lock-in technique, i.e. AC sensor currents, on average the field produced bythe current vanishes. Hence, the bead field measured by the sensor is solely producedby the applied field. The planar Hall sensors’ electronic noise is determined in a setupwhere external noise is reduced. Exactly the low noise level contributes to the high sensorsensitivity. However, the setup can be improved in several ways. A preamplifier for thelock-in amplifier together with increasing the applied field and the sensor current, wouldyield a factor 50-100 on the bead signal-to-noise. This is sufficient to offer single 250 nmbead detection with the studied sensors.

Hereafter, measurements of sensor signal versus antibody concentration are presented.Biotinylated influenza antibodies are immobilized on the surface of a planar Hall sensor.Afterwards magnetic beads with streptavidin surface coating bind to the biotin on theantibodies. The signal is increasing for increasing antibody concentration. Maximumvalue is β = 0.011 (normalized to the applied field), i.e. of the same order of magnitudeas the theoretical estimate.

Finally, the actual influenza immunoassay detection is presented. First fluorescencedetection with directly labelled detector antibodies. Second magnetic detection, where thetotal sandwich assay is performed on top of the planar Hall sensors. The detection principleis thus magnetic beads with influenza antibodies on the surface. S/NC is obtained asthe signal from the sample (S) divided by the signal from the negative control (NC).S/NC = 2 is found for the magnetic immunoassay, and S/NC = 5 for the fluorescenceimmunoassay. The fluorescence assay is optimized, the magnetic is not.

The results obtained with the planar Hall sensors are promising for point-of-care diag-nostics. However, the SU-8 design can be developed further.

Resume

Denne Ph.D. afhandlings titel er Planar Hall sensor anvendt pa influenza immunkemiskdetektion. Afhandlingen omhandler fremstilling, karakteristik og demonstration af planareHall sensorer til anvendelse indenfor influenza immunkemisk detektion.

Malet med dette forskningsprojekt er først og fremmest at designe en magnetisk sensor,som er i stand til at male tilstedeværelsen af sma magnetiske polystyren kugler. Dissekugler varierer i størrelsesorden fra hundrede nanometer til et par mikrometer i diameterog er magnetiserbare i et patrykt magnetfelt. For at kunne integrere denne sensor i enklinisk test, skal sensorprincippet være følsomt og samtidig palideligt. Studier af signal-støjforhold for flere sensortyper, GMR sensorer, spin-ventiler og planare Hall sensorer, pegerpa den planare Hall sensor som bedste valg mht. DC signal-støj forhold. DC malinger vilformentlig være lettere at handtere pa chip-niveau end AC malinger.

Et andet mal med Ph.D. studiet er at demonstrere de planare Hall sensorer i en relevantanvendelse, hvor konventionelle metoder har været utilstrækkelige. Statens Serum Instituthar leveret antistoffer og antigener til demonstrationen af influenza detektion.

En omfattende teoretisk analyse af sensorer og sma magnetiske kugler foretages forat belyse sensorernes anvendelsesmuligheder. Den teoretiske analyse af signalet fra enenkelt 250 nanometer magnetiserbar kugle viser, at de studerede planare Hall sensorer harpotentiale for at male tilstedeværelsen af blot en enkelt kugle. Dette gælder for de underPh.D. studiet fremstillede sensorer. For at male tilstedeværelsen af endnu mindre kugler,kan sensorerne designes til at matche anvendelsen. F.eks. kan 50 nanometer kugler males,hvis det aktive areal er mindre end 1µm × 1µm. Den teoretiske grænse for detektionaf magnetiske kugler som funktion af sensorstørrelse præsenterer gode muligheder for deplanare Hall sensorer. En af grundene er disse sensorers lave interne støjniveau.

Efter studierne med kun en enkelt magnetisk kugle kommer en grundig teoretisk ana-lyse af de renrumsfremstillede sensorer (20µm × 20µm sensorareal) i kombination med250 nanometer store Nanomag-D magnetiske kugler. Netop sensorer og kugler af dissedimensioner benyttes til de efterfølgende eksperimentelle influenza malinger. Et teoretiskoverslag pa signalet fra et monolag af disse kugler er β ≈ 0.024(ξoutside − ξsensor), hvor ξer dækningsgraden i forhold til et tætpakket monolag. β er normaliseret til det patryktefelt og kan saledes sammenlignes med maleresultaterne.

Til sidst i det teoretiske afsnit kigges der pa effekten af at skærme kuglernes positivebidrag med en barriere. Den omtalte barriere kan nemt fremstilles af SU-8, som ogsabenyttes til biokemisk binding af biologisk aktivt materiale. De positive bidrag kan mereeller mindre fjernes ved at placere en sadan barriere ved siden af sensoren. Tilmed kan

v

vi RESUME

indfangning af magnetiske kugler pga. randfelter ved spændingslederne mindskes meddenne barriere. Randfelterne indfanger magnetiske kugler pa omrader, der er af ringebetydning i forhold til malingerne, men biologisk materiale gar tabt ved indfangningen.

Efter de teoretiske analyser følger beskrivelsen af det eksperimentelle arbejde. Herpræsenteres design, fremstilling og karakteristik af planare Hall sensorer. Chip-designetmatcher muligheden for at male tilstedeværelsen af ganske fa 250 nm magnetiske kugler.Chippen er skabt med henblik pa brug i biologiske ekperimenter.

Først præsenteres fremstillingen og resultaterne med nikkelsensorer. De opnaede re-sultater danner grundlag for det videre arbejde med anvendelse af exchange bias til atlave en nem retning i stedet for en nem akse i permalloy baserede planare Hall sensorer.Nikkel sensorerne er fremstillet i MIC’s renrum og udgør et kompromis i forhold til detmagnetiske materiale, da det kun er nikkel, der anvendes i renrummet. Trods deres in-optimale egenskaber beviser nikkelsensorerne imidlertid, at den planare Hall effekt kanbenyttes til at male tilstedeværelsen af magnetiske kugler.

Permalloy sensorerne er optimerede mht. magnetisk materiale og design af en nemretning. Har man en nem akse, vil der være to ækvivalente magnetiske tilstande, somgiver samme signalstyrke med modsat fortegn. Ved at exchange-koble den nemme akse tilen antiferromagnet opnas en nem retning. Saledes har man altid samme udgangspunktfor magnetiseringen i fraværet af et ydre felt. Dette princip anvendes i de optimeredesensorer. Disse sensorer kan male tilstedeværelsen af magnetiske kugler i fraværet af etpatrykt felt, nar der anvendes DC strøm gennem sensoren. Magnetfeltet genereret frastrømmen er nok til at magnetisere kuglerne, hvilket vil være en fordel i en færdig chip.

Ved brug af lock-in teknik, dvs. AC strøm, magnetiseres kuglerne ikke af strømmen.Derfor magnetiseres kuglerne udelukkende af det patrykte felt. Den elektriske støj i deplanare Hall sensorer bestemmes eksperimentelt i en opstilling, hvor udefrakommende støjer minimeret. Netop det lave støjniveau er med til at give sensorerne deres høje følsomhed.Opstillingen kan imidlertid forbedres. Hvis en forforstærker til lock-in forstærkeren benyt-tes, og feltet og sensorstrømmen øges til det maksimale, kan signal-støj forholdet fra enmagnetisk kugle forbedres 50-100 gange. Det ville betyde, at en enkelt 250 nanometerkugle kunne males med de fremstillede sensorer.

Efterfølgende præsenteres malinger af sensorsignal som funktion af antistofkoncentra-tion pa overfladen. Der benyttes biotinyleret influenzaantistof, som fastgøres ovenpa enplanar Hall sensor. Dernæst bindes magnetiske kugler med streptavidin pa overfladen tilbiotinen pa antistofferne. Signalet er voksende som funktion af antistofkoncentration. Denmaksimalt opnaede værdi er β = 0.011 (normaliseret til det patrykte felt). Denne værdiligger indenfor størrelsesordenen af det teoretiske estimat for et monolag.

Afslutningsvis præsenterer afhandlingen egentlig immunkemisk detektion af influenza.Først fluorescensdetektion med direkte mærkede detektorantistoffer. Dernæst magnetiskdetektion, hvor hele sandwichen er lavet ovenpa de planare Hall sensorer. Som detek-tionsprincip er saledes anvendt magnetiske kugler med influenzaantistoffer pa overfladen.Resultatet af disse eksperimenter er S/NC = 2 for magnetisk detektion, hvor S/NCer signalet fra prøven divideret med signalet fra den negative kontrol. For fluorescensmalingernes opnas S/NC = 5, men den immunkemiske sandwich er optimeret for fluor-escens, og ikke for magnetisk detektion.

RESUME vii

Resultaterne for de planare Hall sensorer ser lovende ud for diagnosticeringschips.Desuden kan SU-8 designet udvikles yderligere med forbedring af detektionsprincippetsom resultat.

viii RESUME

Preface

The work presented in this thesis is part of the road towards a magnetic approach inbioassaying. At least for the writer it has been a journey with many challenges, in biologicalas well as physical sciences. The work is not meant as a complete prototype of an influenzachip, only to demonstrate the planar Hall sensor’s capability as sensing principle. Someimprovements will be suggested though they are not experimentally validated due to timelimitations of the project.

This thesis has been written as part of the requirements for obtaining the Ph.D. degreeat the Technical University of Denmark (DTU). The Ph.D. project has been carried outat the Department of Micro and Nanotechnology (MIC) at DTU in the time period July1st 2002 to May 5th 2006.

This Ph.D. project has been a part of the Magnetic Systems research group, MicroElectro Mechanical Systems (MEMS) section, at MIC, and it has been financed by a DTUPh.D. grant. This project has been supervised by Professor Aric K. Menon and AssociateProfessor Mikkel F. Hansen, both employed at MIC.

Part of the work was carried out in the group of Prof. Freitas at Institute of En-gineering of Systems and Computers - Microsystems and Nanotechnology (INESC-MN),Portugal, from September 1st 2003 to December 1st 2003.

Louise Wellendorph EjsingMIC – Department of Micro and Nanotechnology

Technical University of Denmark5 May 2006

ix

x PREFACE

Acknowledgement

The author would like to thank the supervisors of the project for stimulating discussionsand guidance towards the ultimate goal.Professor Aric K. Menon, main supervisor during the main part of the project, whosadly passed away before the completion of the project.Associate Professor Mikkel F. Hansen, co-supervisor and main supervisor during thelast five months of the project.

The author acknowledges the contribution during the external stay in the group ofProf. Freitas at Institute of Engineering of Systems and Computers - Microsystems andNanotechnology (INESC-MN), Rua Alves Redol 9, Lisbon 1000-029, Portugal (www.inesc-mn.pt), especially from Hugo Ferreira and Daniel L. Graham.

Other participation in processing comes from Prof. Hans-Heinrich Gatzen and Dipl.-Ing. Marc Wurz, Universitaet Hannover Institut fur Mikrotechnologie, SchoenebeckerAllee 2, 30823 Garbsen, Germany (www.imt.uni-hannover.de). Additionally by Labora-tory Technician Inger Ninna Hansen, DELTA, Venlighedsvej 4, 2970 Hørsholm, Denmark(www.delta.dk). Electronic Engineer Jan Vasland Eriksen from Danchip is acknowledgedfor constructing the switch box.

For their great help and contribution with the influenza assay, Keld Andresen andLars Peter Nielsen, Statens Serum Institut, Artillerivej 5, 2300 København S, Denmark(www.ssi.dk) are acknowledged together with Sandra Abel Nielsen, Aarhus University,who has assisted in the preface of the assay development. Statens Serum Institut hasprovided the antibodies for the influenza and reference assays.

Additional thanks goes to Associate Professor Martin Dufva for sharing his expertisein bioassays and Associate Professor Ole Hansen for his help with noise investigations,the other members of the group, Kristian Smistrup, Torsten Lund-Olesen, and ChristianDanvad Damsgaard, and the rest of the employees at MIC and Danchip.

Finally, warm thanks to my family and friends, especially my husband Jacob for hisenthusiastic interest in biological science, which has been a great help in understanding ofthis ”other” world.

xi

xii ACKNOWLEDGEMENT

Contents

List of figures xviii

List of tables xix

List of symbols xxi

1 Introduction 11.1 What is a magnetic immunoassay? . . . . . . . . . . . . . . . . . . . . . . . 11.2 How does a magnetic immunoassay work? . . . . . . . . . . . . . . . . . . . 21.3 Why use a magnetic immunoassay? . . . . . . . . . . . . . . . . . . . . . . . 31.4 How is this thesis related to magnetic immunoassays? . . . . . . . . . . . . 41.5 Historical account of the Ph.D. study . . . . . . . . . . . . . . . . . . . . . . 51.6 Outline of thesis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8

2 Magnetic sensors 92.0.1 Requirements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9

2.1 Sensor principles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102.1.1 Anisotropic magnetoresistance . . . . . . . . . . . . . . . . . . . . . 102.1.2 GMR sensor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112.1.3 Spin-valve sensor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 122.1.4 Planar Hall sensor . . . . . . . . . . . . . . . . . . . . . . . . . . . . 132.1.5 Classical sensing principles, the Hall sensor . . . . . . . . . . . . . . 14

2.2 Review of magnetic biosensors . . . . . . . . . . . . . . . . . . . . . . . . . 142.2.1 GMR sensors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 142.2.2 Spin-valve sensors . . . . . . . . . . . . . . . . . . . . . . . . . . . . 142.2.3 Planar Hall sensors . . . . . . . . . . . . . . . . . . . . . . . . . . . . 152.2.4 Hall sensors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15

2.3 Electronic noise in magnetic sensors . . . . . . . . . . . . . . . . . . . . . . 152.3.1 Johnson noise . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 162.3.2 Shot noise . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 162.3.3 1/f noise . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 162.3.4 Noise in planar Hall sensors and spin-valve sensors . . . . . . . . . . 162.3.5 Single bead S/N for planar Hall sensors and spin-valve sensors . . . 17

2.4 Choice of sensor principle . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19

xiii

xiv CONTENTS

3 Theory of planar Hall sensors 213.1 The planar Hall effect . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21

3.1.1 Magnetic energy and expected field response . . . . . . . . . . . . . 233.2 Hysteresis curves . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 263.3 Effects of temperature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28

4 Theoretical sensor signal from magnetic beads 314.1 Magnetic beads . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31

4.1.1 Magnetic characterization of beads . . . . . . . . . . . . . . . . . . . 324.1.2 Bead-bead interaction . . . . . . . . . . . . . . . . . . . . . . . . . . 32

4.2 Sensor and bead . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 344.3 Analytical study of single bead signal . . . . . . . . . . . . . . . . . . . . . 34

4.3.1 Circular sensor geometry . . . . . . . . . . . . . . . . . . . . . . . . 364.3.2 Quadratic sensor geometry . . . . . . . . . . . . . . . . . . . . . . . 364.3.3 Scaling with sensor size . . . . . . . . . . . . . . . . . . . . . . . . . 38

4.4 Single bead detection limits for 20µm× 20µm sensors . . . . . . . . . . . . 384.5 Quadratic sensor and any bead position . . . . . . . . . . . . . . . . . . . . 40

4.5.1 Small sensor compared to bead size . . . . . . . . . . . . . . . . . . . 424.5.2 20µm× 20µm sensor . . . . . . . . . . . . . . . . . . . . . . . . . . . 424.5.3 Monolayer of Nanomag-D beads on 20µm× 20µm sensors . . . . . . 44

4.6 Design of SU-8 layer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 474.6.1 Numerical values for Design 2 . . . . . . . . . . . . . . . . . . . . . . 50

4.7 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51

5 Nickel planar Hall sensors 535.1 Design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53

5.1.1 Sensor appearance . . . . . . . . . . . . . . . . . . . . . . . . . . . . 535.1.2 Sensor material . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 555.1.3 Biologically active material . . . . . . . . . . . . . . . . . . . . . . . 56

5.2 Nickel sensor fabrication . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 565.2.1 Deposition of magnetic Ni films . . . . . . . . . . . . . . . . . . . . . 575.2.2 Photolithography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 575.2.3 Process steps . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57

5.3 Experimental . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 595.3.1 Experimental setup and conditions . . . . . . . . . . . . . . . . . . . 59

5.4 Results and discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 615.4.1 Sensor response vs. angle . . . . . . . . . . . . . . . . . . . . . . . . 615.4.2 Sensor response vs. field . . . . . . . . . . . . . . . . . . . . . . . . . 615.4.3 Bead detection experiments . . . . . . . . . . . . . . . . . . . . . . . 63

5.5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64

CONTENTS xv

6 Exchange biased permalloy planar Hall sensors 676.1 Design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68

6.1.1 Sensor appearance . . . . . . . . . . . . . . . . . . . . . . . . . . . . 686.1.2 Sensor material . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 696.1.3 Biologically active material . . . . . . . . . . . . . . . . . . . . . . . 71

6.2 Clean room fabrication of exchange biased sensors . . . . . . . . . . . . . . 716.2.1 Deposition of magnetic thin films . . . . . . . . . . . . . . . . . . . . 716.2.2 Process steps Batch 1 . . . . . . . . . . . . . . . . . . . . . . . . . . 726.2.3 Process steps Batch 2 . . . . . . . . . . . . . . . . . . . . . . . . . . 73

6.3 Experimental . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 766.3.1 Vibrating sample magnetometry . . . . . . . . . . . . . . . . . . . . 766.3.2 Electrical characterization setup, Batch 2 . . . . . . . . . . . . . . . 766.3.3 Experimental conditions for permalloy sensors . . . . . . . . . . . . 80

6.4 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 816.4.1 Vibrating sample magnetometry of the magnetic films . . . . . . . . 81

6.5 Electrical characterization of permalloy sensors . . . . . . . . . . . . . . . . 826.6 Batch 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83

6.6.1 Bead detection, Batch 1 . . . . . . . . . . . . . . . . . . . . . . . . . 836.7 Batch 2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85

6.7.1 Investigation of noise, Batch 2 . . . . . . . . . . . . . . . . . . . . . 856.7.2 Sensor response vs. field, Batch 2 . . . . . . . . . . . . . . . . . . . . 876.7.3 Investigation of temperature dependence, Batch 2 . . . . . . . . . . 896.7.4 Summary, Batch 2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91

6.8 Discussion of S/N improvement . . . . . . . . . . . . . . . . . . . . . . . . . 926.9 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92

7 Antibody detection 957.1 Immobilizing antibodies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95

7.1.1 SU-8 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 957.1.2 Antibody binding to SU-8 . . . . . . . . . . . . . . . . . . . . . . . . 97

7.2 Antibody binding realized using the planar Hall sensor . . . . . . . . . . . . 997.2.1 Procedure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 997.2.2 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102

7.3 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1087.4 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109

8 Influenza 1118.1 Immunochemistry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1128.2 Fluorescent influenza immunoassay . . . . . . . . . . . . . . . . . . . . . . . 114

8.2.1 Procedure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1158.2.2 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1158.2.3 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116

8.3 Magnetic influenza immunoassay . . . . . . . . . . . . . . . . . . . . . . . . 1168.3.1 Procedure: antibody adsorption to beads . . . . . . . . . . . . . . . 116

xvi CONTENTS

8.3.2 Procedure: planar Hall chip . . . . . . . . . . . . . . . . . . . . . . . 1178.3.3 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118

8.4 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1188.5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121

9 Conclusion and outlook 1239.1 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1239.2 Outlook . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124

9.2.1 Sensor technology . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1249.2.2 Bead design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125

A Analytical expressions 127A.1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127

B Data acquisition and analysis 131B.1 Data acquisition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 131B.2 Data analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 132

C Mathematica code 135

D Publications 141D.1 List of publications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 141

List of Figures

1.1 Sandwich immunoassay. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21.2 Magnetized bead on top of planar Hall sensor . . . . . . . . . . . . . . . . . 3

2.1 GMR stack . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112.2 Spin valve sensor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 122.3 Planar Hall sensor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 132.4 Spectral noise - planar Hall vs. spin-valve sensor . . . . . . . . . . . . . . . 18

3.1 Planar Hall sensor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 223.2 J . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 223.3 Expected sensor response . . . . . . . . . . . . . . . . . . . . . . . . . . . . 253.4 Expected sensor response . . . . . . . . . . . . . . . . . . . . . . . . . . . . 253.5 Normalized film energy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 263.6 Hysteresis loops for uniaxial anisotropy . . . . . . . . . . . . . . . . . . . . 283.7 Hysteresis loops for unidirectional anisotropy . . . . . . . . . . . . . . . . . 28

4.1 VSM measurements of Nanomag-D beads . . . . . . . . . . . . . . . . . . . 334.2 Bead above magnetic sensor . . . . . . . . . . . . . . . . . . . . . . . . . . . 354.3 Circular sensor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 374.4 Quadratic sensor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 374.5 Sensor-bead distance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 394.6 Single bead signal-to-noise for increasing sensor sizes . . . . . . . . . . . . . 414.7 Small sensor compared to bead . . . . . . . . . . . . . . . . . . . . . . . . . 424.8 Varying the bead’s position with respect to the sensor surface . . . . . . . . 444.9 Area division of the evaluated trapezoidal sum . . . . . . . . . . . . . . . . 464.10 Design of the SU-8 capping layer . . . . . . . . . . . . . . . . . . . . . . . . 484.11 Expected signal from a monolayer of Nanomag-D beads covering sensor area 494.12 Expected signal from a monolayer of Nanomag-D beads outside sensor . . . 49

5.1 First nickel planar Hall sensor . . . . . . . . . . . . . . . . . . . . . . . . . . 545.2 Last nickel planar Hall sensor . . . . . . . . . . . . . . . . . . . . . . . . . . 545.3 Magnetic frame . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 565.4 Lithography steps for nickel sensor . . . . . . . . . . . . . . . . . . . . . . . 585.5 Electronic setup for Ni sensor characterization . . . . . . . . . . . . . . . . . 59

xvii

xviii LIST OF FIGURES

5.6 Angle dependence of the Ni planar Hall sensor . . . . . . . . . . . . . . . . 625.7 Signal vs. applied field for nickel planar Hall sensor . . . . . . . . . . . . . . 625.8 Bead detection with nickel planar Hall sensor . . . . . . . . . . . . . . . . . 64

6.1 Final chip, Batch 2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 676.2 Sensor design, Batch 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 686.3 Sensor stacks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 706.4 Lithography steps Batch 2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 756.5 Electronic setup for permalloy sensor characterization . . . . . . . . . . . . 776.6 Experimental setup for sensor characterization . . . . . . . . . . . . . . . . 786.7 Sensor box and temperature control setup . . . . . . . . . . . . . . . . . . . 786.8 VSM measurements for the two sensor stacks . . . . . . . . . . . . . . . . . 826.9 Bead detection Batch 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 846.10 Sensor and setup noise Batch 2 . . . . . . . . . . . . . . . . . . . . . . . . . 866.11 Fit to high field curve - Batch 2 . . . . . . . . . . . . . . . . . . . . . . . . . 886.12 Sensor sensitivity - Batch 2 . . . . . . . . . . . . . . . . . . . . . . . . . . . 886.13 Temperature dependence - Batch 2 . . . . . . . . . . . . . . . . . . . . . . . 906.14 Temperature dependence - Batch 2 . . . . . . . . . . . . . . . . . . . . . . . 90

7.1 Saturation curve for the biotin-streptavidin system of fluorescent and mag-netic beads . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98

7.2 Saturation curve for the biotin-streptavidin system with Cy3-conjugatedstreptavidin . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98

7.3 Procedure for measuring biotinylated monoclonal antibodies with planarHall sensors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99

7.4 Sensor signal as a function of time . . . . . . . . . . . . . . . . . . . . . . . 1007.5 Sensor signal with reference subtracted . . . . . . . . . . . . . . . . . . . . . 1007.6 Stabilization time for planar Hall sensor . . . . . . . . . . . . . . . . . . . . 1037.7 Detection of monoclonal antibodies without reference . . . . . . . . . . . . . 1057.8 Detection of monoclonal antibodies with reference . . . . . . . . . . . . . . 1067.9 Bead coverage . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1067.10 Detection of monoclonal antibodies with reference . . . . . . . . . . . . . . 1077.11 Bead coverage . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107

8.1 Human IgG molecule . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1138.2 Immunochemistry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1138.3 Fluorescence scanning results of influenza immunoassay . . . . . . . . . . . 1168.4 Results of influenza Assay 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . 1198.5 Results of influenza Assay 2 . . . . . . . . . . . . . . . . . . . . . . . . . . . 120

B.1 Relaxation of sensor signal . . . . . . . . . . . . . . . . . . . . . . . . . . . . 134B.2 Normalized off-set corrected sensor signal . . . . . . . . . . . . . . . . . . . 134B.3 Bead detection experiment . . . . . . . . . . . . . . . . . . . . . . . . . . . . 134

List of Tables

2.1 Electronic noise in planar Hall sensors and spin-valve sensors . . . . . . . . 172.2 Single bead signal - planar Hall vs. spin-valve . . . . . . . . . . . . . . . . . 182.3 Single bead signal-to-noise - planar Hall vs. spin-valve . . . . . . . . . . . . 182.4 Signal-to-noise for competitive sensor technologies in magnetic biosensing,

from Freitas, 2004. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19

4.1 Physical properties of magnetic beads . . . . . . . . . . . . . . . . . . . . . 334.2 Scaling of the average bead field . . . . . . . . . . . . . . . . . . . . . . . . 384.3 Single bead S/N and detection limits for a 20µm× 20µm planar Hall sensor 414.4 Area division . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 464.5 Average magnetic field produced by a monolayer of beads . . . . . . . . . . 474.6 Trapezoidal sums, Trp′, of the integrals, Int′ (non-normalized), and the

average magnetic field, β = µ0〈Hx〉/B Eq. (4.31), produced by a monolayerof beads covering the respective area. The design of the SU-8 layer is shownin Fig. 4.10 (2). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50

5.1 Lock-in amplifier settings for Ni sensors. Vout is the output voltage fromthe lock-in amplifier (maximum amplitude) with frequency f . τ is the timeconstant. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60

6.1 Lock-in amplifier settings . . . . . . . . . . . . . . . . . . . . . . . . . . . . 806.2 Theoretical vs. experimental values of planar Hall sensors Batch 1 . . . . . 836.3 Electronic noise in planar Hall sensors Batch 2 . . . . . . . . . . . . . . . . 866.4 Calculated/expected vs. measured values of the planar Hall sensors Batch 2 91

7.1 SU-8 process for 0.5 µm. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 967.2 Physical properties of magnetic beads . . . . . . . . . . . . . . . . . . . . . 977.3 Comparison between magnetic and fluorescent detection systems . . . . . . 1057.4 Estimates of layer coverage . . . . . . . . . . . . . . . . . . . . . . . . . . . 108

8.1 Magnetic immunoassay results . . . . . . . . . . . . . . . . . . . . . . . . . 118

B.1 Lock-in amplifier settings . . . . . . . . . . . . . . . . . . . . . . . . . . . . 131

xix

xx LIST OF TABLES

List of symbols

Symbol Description UnitA Area m2

AMR Anisotropic magnetoresistance effect %aw Atomic weight g mol−1

B Magnetic flux density (magnitude) TB Magnetic flux density (vector) Tc Concentration g l−1

d Orbitale Elementary charge CE Electric field (magnitude) V m−1

E Electric field (vector) V m−1

E(φ) Magnetic energy density J m−3

f Frequency HzF Force vector NH Magnetic field strength (magnitude) A m−1

H Magnetic field strength (vector) A m−1

Ha Effective anisotropy field A m−1

HE Exchange coupling field A m−1

HK Anisotropy field A m−1

I Current AInt Normalized integralJ Current density (magnitude) A m−2

J Current density (vector) A m−2

kB Boltzmann’s constant J K−1

Ku Uniaxial anisotropy constant J m−3

M Magnetization (magnitude) A m−1

M Magnetization (vector) A m−1

M Unit magnetization vector 1MS Saturation magnetization A m−1

n Number of electrons, number of integralsn Surface outward normal vector 1N Noise V Hz−1/2

Nc Number of charge carriers

xxi

xxii LIST OF SYMBOLS

Pl Legendre polynomialsQ Charge Cr, θ, φ Spherical position vectors m, rad, radr, θ, φ Spherical unit vectors 1R Bead radius mRe Electrical resistance ΩR(r) Radial scalar potential T ms OrbitalS0 Sensitivity V T−1 A−1

S/N Signal-to-noise ratioSυ Spectral noise density V2 Hz−1

t Sensor thickness mT Temperature K, oCTrp Trapezoidal sumU Magnetic scalar potential T mv Velocity vector m s−1

V Voltage Vvol Volume m3

w Sensor width mx, y, z Cartesian position vectors mx, y, z Cartesian unit vectors 1α s-d transition probabilityβ Normalized bead field (measured)γH Hooge’s constant∆f Bandwidth Hz∆V Noise V∆V/

√∆f Noise V Hz−1/2

η Actual/saturation magnetizationΘ Debye temperature KΘ(θ) Polar scalar potentialµ0 Magnetic permeability of free space N A−2

ξ Coverageρ,φ Polar position vectors m, radρ, φ Polar unit vectors 1ρ Resistivity tensor Ω mρav Average resistivity Ω mρH Hall resistivity Ω mρ‖ Parallel resistivity Ω mρ⊥ Perpendicular resistivity Ω m% Density kg m−3

τ Lock-in time constant sχg Magnetic mass susceptibility (intrinsic) m3 kg−1

χi Intrinsic magnetic susceptibilityχm Measured magnetic susceptibility

LIST OF SYMBOLS xxiii

ψ Angle between magnetization vectors rad

xxiv LIST OF SYMBOLS

Chapter 1

Introduction

Making things smaller has advantages. As an example is the lab-on-a-chip concept, whereall steps in analyzing a chemical/biological sample are integrated onto a single plastic orsilicon chip. The advantages are numerous. There is need for less reagent, which canbe expensive or hazardous to handle, and need for less sample, which can be difficult toobtain or only exists in small amounts. Moreover, the methodology gives quick resultswith an easy to use dispensable analysis chip.

Using magnetism in lab-on-a-chip systems induces additional benefits. First of all,magnetic actuation and sensing techniques are very well developed since they are usedas the storage/reading basis of computer hard drives - another area of applied physicswhere miniaturization has led to astounding advances. Furthermore, existing knowledgeon material properties is extensive and fabrication techniques are highly developed. Inthe past years magnetic sensing has been applied to bioassaying, with possible favorableoutcome compared to existing techniques. The biomolecule is labelled with a magneticbead and its presence detected via the label. This way of detecting a biomolecule is oneof the only existing methods which are sensitive enough to detect a single molecule.

The most promising results of magnetic sensing in bioassaying has been obtained usingsensors based on the giant magnetoresistive (GMR) effect of multilayered structures. Thisthesis demonstrates that sensors based on the planar Hall effect give comparable resultswith a sufficiently favorable signal-to-noise (S/N) ratio to detect a single magnetic label.The planar Hall sensors are easily fabricated and can be integrated with fluidics to accountfor the sensing part of a lab-on-a-chip for bioassaying.

1.1 What is a magnetic immunoassay?

An immunoassay imitates a higher organism’s reaction towards an infection by a for-eign microorganism, which expresses antigens. The antigens are recognized as foreign bythe host’s immune system. The immune system responds to the presence of antigens byproducing specific antibodies against a selection of antigens expressed by the foreign mi-croorganism. It is this specific reaction of antibodies with their corresponding antigen,like a key in a lock, which is used in immunoassays. In a sandwich immunoassay (Fig. 1.1)

1

2 CHAPTER 1. INTRODUCTION

Figure 1.1: Sandwich immunoassay. The capture antibody is immobilized on the surface.Then the antigen is attached to the immobilized antibody and reaction with the detectorantibody can take place. In this assay sketch the detector antibody is labelled with biotin.

the researcher is looking for certain antigens, e.g. A and B. An antibody against A is im-mobilized on a solid substrate (spot A) and an antibody against B at a different position(spot B). These are called capture antibodies. Then the sample is added to the substrate.If the sample contains antigen A, it will bind to the antibody of spot A, and if the samplecontains antigen B, it will bind to the antibody of spot B. Then a mixture of antibodiesagainst antigen A and B are added to the substrate. These antibodies are called detectorantibodies and they can be marked prior to the experiment; in the figure they are markedwith biotin. Biotin binds specifically to streptavidin. Streptavidin is a protein, which cancarry a color dye or a fluorescein molecule. Then from the position of colored spots orfluorescent spots, the researcher is able to determine what antigen, A or B, was originallyin the sample.

If streptavidin-coated magnetic beads are used instead of the labelled streptavidin,with color or fluorescent dye, the assay will be marked magnetically. This is the principleof a magnetic immunoassay.

1.2 How does a magnetic immunoassay work?

Recapitulate the sandwich immunoassay: capture antibodies are immobilized on a solidsubstrate, antigen reacts with the capture antibodies, and biotin labelled detector anti-bodies bind to the captured antigen. In a magnetic immunoassay the capture antibodiesare immobilized on top of a magnetic sensor. Following antigen addition, biotin labelled

1.3. WHY USE A MAGNETIC IMMUNOASSAY? 3

Figure 1.2: Schematic illustration of a magnetized bead on top of a planar Hall sensor,top view. The sensor is green and the magnetized bead is shaded grey - its magnetizationillustrated as an arrow. The field lines just below the bead are opposite to the appliedfield.

detector antibodies are introduced, and streptavidin coated magnetic beads bind to thebiotin. Finally the system is flushed to remove unbound beads. The outcome is a magneticbead attached on top of a magnetic sensor, if and only if the antigen is contained in thesample. The positive situation is sketched in Fig. 1.2, where a magnetic bead is shownon top of a magnetic sensor. The presence of the bead alters the field registered by thesensor in comparison to the absence of the bead, see sketched field lines.

1.3 Why use a magnetic immunoassay?

Ordinary influenza, the viral disease that returns approximately every year, results in in-creased death rates. However, a pandemic, such as the 1918 Spanish Flu, will be lethalto a considerable part of the World’s population. Antiviral drugs have been developedagainst influenza, and an early treatment decreases the duration of illness and alleviatesthe symptoms. These antiviral drugs are designed to help the body’s own immune systemfighting the infection, and although it is not necessarily fully effective, it might help indi-viduals survive. In order to have an effect, however, it is essential that the drug is takenas soon as possible after the infection, compulsorily within 48 hours. Therefore, a quickand reliable diagnostic test is needed.

Today, reliable testing is done with polymerase chain reaction (PCR). PCR requiresspecial equipment and skilled personnel, hence using the technique is expensive and time


consuming in the sense that samples taken at the doctor’s must be transported to specialplaces and the test results must be communicated back to the consulting room. A varietyof fast tests exist, based on immunochemistry, but they are not reliable. The reason forthe failure of these immunochemistry based tests is partly a very low concentration of viralprotein, the antigens, and partly because the samples contain enzymes that break downthe antibodies the test is based on. The last issue is often solved by diluting the sample,but unfortunately this aggravates the first problem.

Using magnetic beads, we have the option of capturing the sample antigen by anti-bodies attached to the magnetic beads. On the surface of the beads the antibodies areimmobilized in advance to the capturing event. These beads are suspended in the sampleand antibodies attached to the beads are allowed to react with the antigens on the surfaceof the sample molecules. Afterwards, the antigens are held back with magnetic forceswhile the remaining contents are washed away. This solves both issues in one step. Theantigen is held back, when diluting the enzymes from the sample.

Next step is to detect the magnetic bead in a magnetic immunoassay, which onlybinds to the capture antibodies on top of the sensor if an antigen is attached to the bead.By detecting the carrier, one avoids elution of the antigen into new buffer, which woulddecrease the antigen concentration and thus also the sensitivity of the assay. Unspecificallybound beads can be removed from the sensitive region of the sensor with magnetic forces.This would eliminate the need to wash the probe, which can be tedious especially forpoint-of-care diagnostic tests. The readout from the sensor is electronic, which means thatcontrol systems and calibration can be integrated on the chip, and neither laser scannernor PCR equipment is needed. Magnetic sensors such as the planar Hall sensor are verysensitive and can potentially detect a single bead and thus potentially the presence of asingle antigen.

1.4 How is this thesis related to magnetic immunoassays?

In the following chapters, the application of the planar Hall sensor to magnetic immunoas-says will be addressed. The planar Hall sensor is an attractive choice for the sensing partof the significant challenge that a magnetic immunoassay presents. It turns out that in acomparison of competing technologies’ signal magnitudes and noise levels for DC detec-tion, the planar Hall sensor outperforms its competitors. The investigation is addressedin terms of the single bead signal-to-noise level.

During the research leading to this thesis, planar Hall sensors of various design andbehavior have been fabricated and characterized. A large effort has been invested in designconsiderations and optimization, and these issues are addressed in different chapters, whichfollow the chronological development of the sensor design and performance. First, thenickel planar Hall sensors are presented. Though not optimal with respect to design andmaterial, they represent an important first step in optimizing these aspects in the improvedversion of exchange biased permalloy sensors. During this period essential parts of theelectrical characterization setup are constructed and put together and the extended effortsin noise suppression initialized. The nickel sensors demonstrate the use of the planar Hall

1.5. HISTORICAL ACCOUNT OF THE PH.D. STUDY 5

effect as a magnetic sensing principle, and bead detection is demonstrated.The results obtained with the nickel sensors guide the design and fabrication of the

permalloy sensors. The new sensors are exchange-coupled to an antiferromagnetic materialin order to induce a magnetic easy direction instead of only an easy axis, as is the caseof nickel sensors. The exchange biased permalloy sensors are fabricated in collaborationwith INESC-MN in Portugal and two batches are produced. Batch 1 is a bottom-pinnedstack and Batch 2 is a top-pinned stack. Batch 1 is used to demonstrate bead detectionin collaboration with INESC-MN, and Batch 2 is used for the influenza immunoassayexperiments.

Resuming the efforts in noise suppression has lead to approximately 100 times less noisethan connected to the experiments performed at INESC-MN. The planar Hall sensors ofBatch 2 are then characterized magnetically, electrically, and the noise is determined alongwith the temperature dependence of the signal. Batch 2 sensors are applied to the influenzaimmunoassay. First detection of monoclonal antibodies is demonstrated, where the signalas a function of concentration is determined. Second successful detection of an influenzasandwich immunoassay is carried out with the planar Hall sensors.

On the biochemistry side of the thesis, a new immobilization method is developed incollaboration between people at MIC. DNA and antibodies are found to attach readily tothe surface of SU-8, a photosensitive polymer widely used in MEMS fabrication. A protocolfor the influenza sandwich assay on SU-8 is developed and test results with fluorescentlylabelled detector antibodies are presented.

1.5 Historical account of the Ph.D. study

1. During a literature survey for very sensitive magnetic sensors, an article by Mon-taigne et al. [1] was encountered. This article describes a magnetic field sensorwith nanotesla resolution based on the planar Hall effect of permalloy. The interestfor this planar Hall sensor had been rather low. Such sensors had not been con-sidered for read heads in computer hard drives, where GMR sensors are preferred.Moreover, they had not been considered for bead detection either. The suggestionby Montaigne et al. is to use the planar Hall sensor as a micro-compass. A theo-retical evaluation of the sensing principle (presented in Chapter 3) has lead to theunderstanding that this type of sensor is applicable to magnetic bead detection.

2. After theoretical validation of the sensing principle, an experimental validation wasneeded. The fabrication of nickel planar Hall sensors was carried out in MIC’s cleanroom, and since the only ferromagnetic material available was nickel, nickel wasused as active sensing material. The results from this investigation proved that theplanar Hall sensing principle was sufficient to detect 2.8 µm Dynabeads, resulting ina conference proceeding [2]. During the work with nickel sensors, students at MICbuilt a set of Helmholtz coils [3], which were calibrated during this project. Thecalibration is 2.8 % too low compared with the recent, and more reliable, calibration.Hence, the results with the nickel sensors refer to a slightly different coil calibrationthan more recent results. For the electrical measurements and current control, a


lock-in amplifier was available but no high precision current source for sensors orcoils (the electrical setup of Fig. 5.5 was used). Insights essential for the design ofexchange biased permalloy sensors was obtained from the nickel sensor results. 1:Need for an easy direction instead of an easy axis. 2: Need for an exchange fieldto enhance the effective film anisotropy of permalloy as well as to create the easydirection.

3. Since it is not possible to deposit the preferred materials in MIC’s clean room, thestudied exchange biased permalloy sensors have been fabricated in collaborationwith INESC-MN, Portugal. First Batch 1, the bottom-pinned stack, was fabricatedcompletely in the laboratories in Portugal, and all characterization was made usingtheir facilities. In collaboration detection of 2 µm Micromer-M beads and 250 nmNanomag-D beads was demonstrated [4, 5]. The magnetic characterizations per-formed on MIC’s VSM revealed an ”extra” loop in the M vs. H measurements dueto a magnetic seed layer for MnIr growth. This is undesirable and was subsequentlyeliminated using a top-pinned stack for Batch 2. The bead detection experimentswith Batch 1 suffered from a considerable amount of electronic noise. The setupwas without electrical screening and with ordinary unshielded wires. Encouragedby the potential for noise reduction, a substantial effort has been put into avoidingelectronic noise in the author’s own setup.

4. The fabrication of sensor Batch 2 has been carried out at four different places. Themagnetic films were deposited at INESC-MN, the photolithography was made atMIC, and etching of the sensor structures at Universitat Hannover. The subse-quent processing was done at MIC apart from wire bonding, which was done atDELTA. Due to time constraints no further sensor batches were made, though pos-sible improvements were discovered after finishing processing. The best example of adesign parameter that could have been improved, is the SU-8 layer for immobilizingbiomolecules. In the processed design, only the sensitive area of the sensor is coveredwith an SU-8 pad. The reason behind this choice is that biomolecules are supposedto attach only on top of the sensor. This would avoid wasting sample molecules atspaces, which cannot be sensed by the sensor. Unfortunately the molecules bindas easily to the residual monolayer adjacent to the SU-8 pad as to the pad itself.In this design, competitive positive and negative signals result in reduced averagesignals. This was not known at the time of processing, and the choice was at thattime justified.

5. The setup for electrical characterization was built after completing the fabrication.At first the old setup used for the nickel sensor (Fig. 5.5) was available. A KeithleyModel 6221 current source was borrowed for the Helmholtz coils, and the sensor cur-rent was controlled by the lock-in amplifier. A 5 kΩ low-noise resistor was connectedin series with the sensor circuit to implement a better sensor current control. In thissetup, the noise characterization of Batch 2 was made. Noise reduction was obtainedby shielded coax cables and a shielded sensor box. The measured electronic noise isapproximately 100 times less than that measured at INESC-MN.

1.5. HISTORICAL ACCOUNT OF THE PH.D. STUDY 7

6. The experimental setup was gradually rebuilt (resulting in the setup of Chapter6, Fig. 6.5 and Fig. 6.6). One significant improvement is the introduction of twonew Keithley Model 6221 high precision current sources for current control of theHelmholtz coils and the sensor current. Since the planar Hall sensor is temperaturesensitive, the sensor box has been rebuilt with thermal insulation and temperaturecontrol. Additionally, a switch box was constructed to facilitate measuring all threesensors on a chip. The possibility of subtracting the reference sensor signal from thesignal from the sensor measuring biological activity is thus introduced. Incorporatedon the chip is sensor A for the sample, sensor B for the electrical reference, andsensor C for the negative control. As the three sensors are serially connected, theyexperience the same applied current.

7. In the final setup, the characterization of the planar Hall sensors of Batch 2 has beenperformed. Finally, theory and experiment compare reasonably. The temperaturedependence of the sensors was measured and found reasonably linear. Then theprocedure for measuring antibodies immobilized on top of a sensor was developed(the resulting procedure is presented in Appendix B).

8. The presented reference system is biotinylated monoclonal antibodies bound to strep-tavidin coated magnetic beads. The bead signal as a function of antibody concen-tration is determined. A general trend of increasing signal with increasing antibodyconcentration is obtained. However, the signal from the bead layer is positive in-stead of negative as expected from the theoretical estimates performed earlier. Firstphysical alteration of the sensors during chemical processing was assumed, but thesignal remained positive. Ultimately, when calculating the signal from a monolayerof beads with the chip’s actual SU-8 configuration, it was discovered that the signalcould indeed be positive (Chapter 4). The reason is a tradeoff between positive andnegative field contributions. The field from beads adjacent to the sensor gives a pos-itive contribution, and beads placed directly above the sensor’s sensitive area give anegative contribution. The relative bead coverage determines whether the outcomeis a positive or negative signal.

9. During the investigation of the positive sign from the monolayer of magnetic beads,it was found that another design of the SU-8 layer would be preferable. Namely anegative image of what was actually fabricated. If the positive contribution frombeads placed outside the sensitive area of the sensor were shielded by a thicker SU-8layer, the signal would not only become negative as it should, but bead capture atfringing fields from the voltage leads could also be reduced. Unfortunately, this wasonly found after the puzzle about the ”wrong” sign was solved, which occurred at avery late stage of the Ph.D. study, so time did not permit to test it experimentally.

10. Finally, a successful influenza immunoassay on the planar Hall sensor chip was per-formed (Chapter 8). Taking the design and other issues into account, the results arepromising for the planar Hall sensors.


1.6 Outline of thesis

The aim of this work is to fabricate magnetic sensors and demonstrate their feasibility asbiosensors in point-of-care diagnostics of influenza infection.

Outline of chapters

1. Chapter 1 introduces the magnetic immunoassay.

2. Following a description of the physical sensing principle for various magnetic sen-sors in Chapter 2 comes a short review of completed experiments towards magneticbioassaying in other research groups. Chapter 2 concludes the choice of the planarHall sensor based on the signal-to-noise level compared to existing sensor types.

3. Theoretical background of the planar Hall sensor is the content of Chapter 3.

4. More theory underlying the study of single bead signal as well as the monolayersignal is presented in Chapter 4. Additionally, various SU-8 designs are consideredwith respect to the theoretically expected monolayer signal.

5. Chapter 5 deals with the clean room fabrication of nickel planar Hall sensors alongwith the characterization and bead detection results using this sensor batch.

6. Chapter 6 presents the permalloy sensor fabrication and characterization of Batch 1and 2. Following a description of the experimental setup used for characterizationare first the magnetic then the electric characterizations including sensitivity, noise,and temperature investigations.

7. The signal produced by magnetic beads attached to different surface concentrationsof biotinylated monoclonal antibodies represents the baseline for detection experi-ments with immunoassays. Chapter 7 presents the procedure of immobilizing anti-bodies on SU-8 placed on top of the sensor, then follows the baseline studies withplanar Hall sensors.

8. Chapter 8 deals with the actual influenza immunoassay realized using the planarHall sensors. Three sensors are used, where one is a reference used to stabilize thesignal from the other two, the second is used for the negative control, the third isreserved for the sample.

9. Chapter 9 concludes the thesis.

Chapter 2

Magnetic sensors

When deciding on a sensor for a magnetic immunoassay, the sensing principle and thesignal that this principle produces is important. However, the detection limit of thesensor also depends on the noise generated during measurements. This chapter describesthe various sensor principles along with a discussion of electronic noise. The planar Hallsensor offers the best DC signal-to-noise, S/N , from a single magnetic bead compared tothe other types of sensors.

The chapter includes the requirements for magnetic sensors in magnetic bioassaying, adescription of the physical sensing principles considered, and a review of results obtainedprior to this thesis. Additionally the electronic noise in magnetic sensors is evaluatedleading to the choice of the planar Hall sensor for magnetic bioassaying.

2.0.1 Requirements

Some requirements for a magnetic bead sensor are the geometry of the sensing principle,the range of fields the sensor is sensitive to, and the linearity of the sensor. If the sensorgives a linear response to the applied field, the sensor measures an average of the beadfield over the entire sensing area.

First the sensing geometry of a magnetic field sensor placed in the xy-plane of acartesian coordinate system is considered. If the sensing principle is based on the Lorentzforce, the bead must be magnetized perpendicularly to the sensing plane, i.e. along z. Ifthe sensing principle is based on magnetoresistance, the bead can be magnetized parallelto the sensing plane, i.e. along x.

A magnetic dipole moment oriented along z produces field lines that will enter and leavethe sensor surface in equal amounts. This is due to flux closure. Thus the z-componentof the bead field averages to zero when the area is large compared to the bead size. Asensor area has to be small compared to the bead for this average to approach a valuefor detection. For nanobeads, strong constraints on the fabrication procedure must beexpected.

When the bead is magnetized along x, a rather large negative field component justbelow the bead dominates the dipole field. This component does not average to zero whenthe sensor area is large compared to the bead size, hence larger sensor areas can be used.

9

10 CHAPTER 2. MAGNETIC SENSORS

However, the single bead signal scales inversely with sensor size cubed (see Chapter4), and for single bead detection an area, which is small compared to the bead, would beideal. In biological sensors, we cannot rely on a single molecule to place itself just on topof a very tiny sensor, and in this context single bead detection is meaningless. Here weprefer an area that probes an average of the sample, and it must be large compared tothe bead. For biological experiments a statistical bead count is important, thus only thesensing geometry along x can be used. If room for at least one thousand beads is needed,the sensor area becomes 1000× the bead size. Going to larger sensor areas clarifies theconstraint of the sensing geometry, only the in-plane magnetization of beads can be used,not the out-of-plane magnetization which would mainly measure beads situated at theedges of the sensor.

Superparamagnetic beads are designed to avoid remanent magnetization and hencethey must be magnetized by an external field. Therefore the sensor must be able todistinguish a very small field component compared to the background magnetic field.Chapter 4 explains that this component is approximately 1 % of the applied field. Hencethe sensor should have a signal-to-noise ratio that enables detection of ∆B = 10 nT in aflux density of B = 1 mT (a typical value for MR based sensors). It should also have awell-defined field response up to the applied B = 1 mT, or more generally up to 100 timesthe detection limit ∆B.

2.1 Sensor principles

This section describes the competing sensor technologies for magnetic bioassaying. Theseinclude GMR sensors, spin-valve sensors, and the planar Hall sensors, which have all beenconsidered as possible sensing principles. The signal produced by the sensor in an appliedfield is presented.

2.1.1 Anisotropic magnetoresistance

Magnetoresistance (MR) is present in a ferromagnetic conductor [6]. MR stems from thespin-orbit coupling between electrons and magnetic moments of the lattice atoms and isthus quantum mechanical in origin as opposed to ordinary galvanomagnetic effects thathave a classical origin in the Lorentz force. Extraordinary effects are created by themicroscopic part of the flux density (µ0M) whereas the ordinary effects are created bythe macroscopic part (µ0H). The GMR effect, the spin-valve sensor, and the anisotropicmagnetoresistance (AMR) effect arise from the MR effect. The Hall effect arises from theordinary galvanometric effect.

Anisotropic magnetoresistance (AMR) also stems from the spin-orbit coupling, but itis anisotropic. AMR depends on the crystal structure of the material and the spin of theatoms situated in the crystal. It depends on the domain structure of the ferromagnet.In general the resistivity is larger if the current is applied parallel to the magnetizationthan if the current were applied perpendicularly to the magnetization (ρ‖ > ρ⊥). Thisdifference between the two states of magnetization is measured as the AMR effect, AMR =

2.1. SENSOR PRINCIPLES 11

Figure 2.1: GMR stack (Fig. 3 in White, 1994). Schematic drawing of the conductionin a multilayered GMR film. Different scattering produces different resistances in theantiparallel (2a) and parallel (2b) cases of magnetization alignment, here shown throughthe mean free path, which is larger in the parallel case.

(ρ‖ − ρ⊥)/ρav, where ρav is the average resistivity of the ferromagnetic material. It is theanisotropic magnetoresistance effect that is responsible for the planar Hall effect.

2.1.2 GMR sensor

The GMR effect reported by Baibich et al. [7] is based on antiferromagnetic coupling ofFe layers through thin Cr separation layers, called spacers. Electrons travelling througha magnetic layer in a GMR stack, as the one shown in Fig. 2.1 (Fig. 3 in Ref. [8]),have spins aligned according to the magnetization of the ferromagnetic layer. In theband structure of some ferromagnetic materials, the Fermi surface intersects the band ofone spin component and not the other, hence only electrons of this spin are available tocarry current. The result is a spin polarized current, which can only be accepted intobands of identical spin in adjacent layers. The resistance then strongly depends on therelative orientation of magnetization in the closely spaced ferromagnetic layers. Variousspacing layers show different results. At specific Cr layer thicknesses the coupling isantiferromagnetic (Fig. 2.1, 2a), at other layer thicknesses the coupling is ferromagnetic(Fig. 2.1, 2b). The magnetoresistive effect observed for the antiferromagnetic couplingbetween layers reported by Baibich is up to 50 %, an order of magnitude higher thanprevious experiments with magnetoresistive materials.

The GMR effect is described by the change in resistivity, ∆ρ(ψ), normalized to the


Figure 2.2: Spin valve sensor reported by Ferreira et al, 2003. Schematic drawing showingthe configuration of magnetic moments in the sensing layer (free layer) and the pinnedlayer. The easy axis is indicated with a dotted line, and the external field is appliedperpendicularly to this.

average resistivity, ρ, [6]

∆ρ(ψ)ρ

=(

∆ρ

ρ

)

(GMR)

1− cosψ

2(2.1)

where ψ is the angle between the magnetization in the two sets of layers, and the (∆ρρ )(GMR)

is the maximum attained value, the GMR coefficient. For antiferromagnetically coupledlayers (corresponding to Fig. 2.1, 2a) the stack can be used as a magnetic field sensor. Thefield, H, is applied along the hard axis of magnetization, which corresponds to H appliedacross the layers in the figure. The field dependence of Eq. (2.1) is to a first approximation[6] (

∆Re(H)Re

)

GMR

=(

∆ρ

ρ

)

(GMR)

[1−

(H

Ha

)2]

(2.2)

where ∆Re(H)/Re is the normalized resistance change, and Ha is the effective anisotropyfield.

2.1.3 Spin-valve sensor

The working principle of the spin valve sensor is sketched in Fig. 2.2 (Fig. 1 in Ref. [9]).When placed in an external magnetic field the magnetization of the free layer is rotatedaway from the easy axis (dotted line in the figure) and gives rise to a change in resistancethrough the sensor, ∆Re, between the two contact pads. The resistance depends, as in thecase of GMR sensors, on the relative orientation of the magnetization between the pinnedand the free layer.

With the easy axis of the sensing layer perpendicular to the reference layer, i.e. thepinned layer, as shown in Fig. 2.2, the field dependence of the sensor is [6]

(∆Re(H)

Re

)

SV

=∆ρ

ρ

(1−H/Ha)2

(2.3)

2.1. SENSOR PRINCIPLES 13

Figure 2.3: Left: Geometry of the planar Hall sensor. The magnetization vector, M, hasan angle φ with respect to the x-axis. M lies in the xy-plane, the current, I, is along y,and the voltage, V , is measured along x. The sensor cross has the width w and thicknesst. Right: Cross sectional view of the sensor stack. Layer thicknesses are in A. The sensingis governed by the NiFe layer, which is pinned by the MnIr layer.

where H is applied perpendicularly to the easy axis, and Ha is the anisotropy field of thefree layer. The remaining parameters have been introduced for the GMR sensor. Thevoltage drop, V , in an applied field becomes

VSV = −12

I Re∆ρ

ρ

H

Ha(2.4)

where I is the current through the sensor. The spin-valve sensor is linear for applied fieldsup to Ha.

2.1.4 Planar Hall sensor

The planar Hall effect, Fig. 2.3, is generated when the magnetization, M, is rotatedwith respect to the current, I. The origin of the planar Hall effect is the anisotropicmagnetoresistance effect, which is different from the origin of the Hall effect. Namingthis effect planar Hall is bound to cause misunderstanding, but nevertheless, the term iswell known and abundant in the literature. The connection between the Hall effect andthe planar Hall effect is solely geometrical. The Hall effect is sensitive to magnetic fieldsapplied perpendicularly to the sensor plane, and the planar Hall effect is sensitive to fieldsapplied parallel to the sensing plane, hence the name planar.

The mathematical expression of the planar Hall sensor signal produced by an appliedfield is

VPH = I∆ρ

ρ

ρav

t

H

Ha(2.5)

in which Ha is the effective anisotropy field. The planar Hall sensor is linear within 2 %for H < 0.23Ha.

Details of the working principle behind the planar Hall effect is given in Chapter 3,where Eq. (2.5) is derived.


2.1.5 Classical sensing principles, the Hall sensor

The Hall effect arises from the Lorenz force deflection of charge carriers, with charge Qand velocity v, through a medium. The force, F, is directly proportional to the crossproduct between the carrier velocity, v, and the magnetic flux density, B.

F = Q(v ×B)

This gives rise to a Hall resistivity, ρH, [6]

ρH =B

newhere n is the density of free charge carriers of elementary charge e.

The Hall sensor’s sensing geometry is suboptimal for magnetic bioassaying as discussedin the requirements subsection. Only in the case where the sensor size approaches the beaddiameter, the flux through the sensor differs and a Hall signal can be obtained. Anotherway to break the symmetry is to apply a time-varying magnetic field, see Besse et al. [10].

2.2 Review of magnetic biosensors

Having described the physical detection principles, this section reviews the detection ex-periments performed using the magnetic sensors. Only experiments relevant to magneticbioassaying or to this thesis are included.

Existing magnetic field sensors used for biomaterial detection by magnetic labelling areGMR sensors, spin-valve sensors, and Hall sensors. Planar Hall sensor have not previouslybeen used as biosensors or to detect magnetic beads. Below is presented a review ofbiomolecular detection experiments performed using various sensor types.

2.2.1 GMR sensors

Giant magnetoresistance (GMR) was discovered in 1988 by Baibich et al. [7], who pub-lished MR ratios (the change in resistivity due to an applied field) an order of magnitudelarger than seen before. The GMR effect was hereafter rapidly exploited first as a mag-netic field sensor in 1994 [11] second by the recording industry to enhance the sensitivityof hard disk drives (IBM 1997). In 2000 Edelstein et al. published the detection of biolog-ical warfare agents using a bead array counter (BARC) biosensor, which was essentiallyan array of GMR sensors [12]. Later the same group demonstrated detection of DNAhybridization using similar BARC chips [13].

Recently, nanometer sized GMR sensors were fabricated and characterized [14]. Thehereby obtained signal-to-noise level of the sensors was promising with respect to single100 nm bead detection but no actual detection experiments were performed.

2.2.2 Spin-valve sensors

Spin valve sensors have been demonstrated as biosensors by Prof. Freitas’ group (INESC-MN, Portugal, www.inesc-mn.pt) [9, 15, 16, 17, 18]. In Ref. [15] detection and manipula-tion of single 2 µm magnetic labels, Micromer-M, were demonstrated using on-chip current

2.3. ELECTRONIC NOISE IN MAGNETIC SENSORS 15

lines. In Refs. [9] and [16] streptavidin coated beads were bound to a biotinylated sensorsurface and detected with the spin valve sensor. Recently, Graham et al. demonstrateddetection of DNA hybridization events using their spin-valve chips [18]. Additionally, sin-gle bead detection of 2.8 µm labels, Dynabeads M-280, with spin-valve sensors has beenperformed by Li et al. [19].

2.2.3 Planar Hall sensors

In 2000, when publishing their work on sensors based on the planar Hall effect of permalloy,Montaigne et al. [1] suggested the possibility of using this type of sensor as a micro-compass. Their permalloy sensors were saturated in the Earth’s magnetic field, i.e., notfeasible as bead detectors since saturation of the sensor would occur prior to magnetizationof the beads.

Neither bead detection nor magnetic bioassaying has been demonstrated with planarHall sensors prior to this thesis.

2.2.4 Hall sensors

In 2002 Besse et al. [10] published a miniaturized silicon Hall sensor capable of detectinga single microbead of diameter 2.8 µm (Dynabeads M-280). In order to do so they mag-netized the bead with a constant external field and added an AC field either parallel tothe constant field or perpendicular to the constant field. Detection of the bead via theapparent susceptibility (HAC parallel to the applied field) or the second-harmonic (HAC

perpendicular to the applied field) was successful in both cases.An InAs quantum-well Hall sensor was also proven capable of detecting a single bead

[20], this time 1.2 µm in diameter.

2.3 Electronic noise in magnetic sensors

This section deals with electronic noise. After a description of three sources of intrinsicnoise, the noise in planar Hall sensors and spin-valve sensors is compared.

Electronic noise in sensors arises from various sources. A very well-known source ofnoise is the Johnson noise, studied by Johnson in 1927 [21, 22], which has its origin inthermal agitation of the charge carriers in a resistor. Another type of noise, shot noise, isrelated to the current passing an energy barrier. The shot noise arises from the quantumnature of charge carriers, and is only expressed when the current fluctuates. Another sortof noise is the 1/f noise, which is of unknown origin but depends strongly on frequency.Hence the name 1/f . Noise, ∆V/

√∆f , is measured in units of volt per square root of

Hertz (V Hz−1/2). Alternatively, the experimental noise, ∆V (standard deviation on thevoltage), is measured in units of volt for a specific bandwidth, ∆f .


2.3.1 Johnson noise

Any electronic device, that is sensor, amplifier, serial resistor, etc. exhibits Johnson noise.A resistor has thermal noise according to Eq. (2.6).

∆VJ/√

∆f =√

4kBTRe (2.6)

where kB is Boltzmann’s constant, T the temperature, Re the electric resistance, and ∆fthe measurement bandwidth.

2.3.2 Shot noise

When an alternating current (IAC) is used, the current varies with time and is thus subjectto the quantum mechanical distribution of the charge carriers that governs shot noise. Themathematical expression of shot noise is [23]

∆Ish =√

2eI∆f (2.7)

∆Vsh/√

∆f = Re ·∆Ish/√

∆f

∆Vsh/√

∆f = Re

√2eI

where e is the elementary charge, and the rest of the parameters are presented previously.

2.3.3 1/f noise

A tricky sort of noise is the 1/f noise, named so due to its spectral dependence. Atpresent, the origin of the 1/f noise is unknown, perhaps due to its ubiquitous nature. Themathematical expression of the spectral noise density, Sυ(f), is [24]

Sυ(f) = γHV 2(Ncf)−1 (2.8)

∆Vf/√

∆f =√

Sυ(f) = V√

γH(Ncf)−1

where f is the frequency, γH is Hooge’s constant (dimensionless), V the voltage drop, andNc is the number of charge carriers in the resistor.

Hooge’s constant has been found to be γH = 10−3-10−1 for permalloy in varyingmagnetic fields, where the magnitude depends on the internal domain structure [25]. Forapplied fields corresponding to B = 0.5 mT, Hooge’s constant is γH ≈ 10−2 [25]. Thenumber of charge carriers can be estimated from the volume, density, and the atomicweight of the sensor material. In the following analysis, one carrier per atom is assumed.

2.3.4 Noise in planar Hall sensors and spin-valve sensors

An estimate of the electronic noise in a typical planar Hall sensor and spin-valve sensorof identical geometry yields the values presented in Table 2.1. The 2µm × 2µm activesensing area is chosen for detection of a single magnetic bead with a diameter of 2 µm.The parameters for the spin-valve sensor, including γH, are from Ref. [26]. γH for theplanar Hall sensor is from [25], the remaining parameters will be given in Chapter 6.

2.3. ELECTRONIC NOISE IN MAGNETIC SENSORS 17

t (nm) Re (Ω) I (mA) γH Nc NJ ( nV√Hz

) Nsh ( nV√Hz

)PH 20 10.85 1 0.01 7 · 109 0.4 0.2SV 10 21.70 1 0.10 3.5 · 109 0.6 0.4

Table 2.1: Electronic noise, N , in planar Hall sensors (PH) and spin-valve sensors (SV).Both sensors have 2µm × 2µm active sensing area. Re = ρav/t is the resistance of thesensor material with thickness t. I is the sensor current, γH Hooge’s constant, Nc thenumber of charge carriers. NJ: Johnson noise, Nsh: shot noise.

Fig. 2.4 shows the sensor noise as a function of frequency for both sensors. The whitenoise is the sum of Johnson noise and shot noise. For the planar Hall sensor the whitenoise is approximately half of that for the spin-valve sensors. However, the 1/f noisedominates the low frequency part of the spectrum, and the planar Hall sensor exhibits 10times less 1/f noise than the spin-valve sensor. The figure indicates a lower noise levelof the planar Hall sensor than of the spin-valve sensor, which is especially pronounced forlow frequencies, i.e. in DC measurements.

2.3.5 Single bead S/N for planar Hall sensors and spin-valve sensors

To make a signal-to-noise estimate of the planar Hall sensor and the spin-valve sensor, thesignal produced by a single 2 µm bead is studied. The signal produced by a bead centereddirectly on the surface of a 2µm× 2µm active sensing area is

〈Hbead〉 = − χmH/3(1 + w2)(1 + 2w2)1/2

(2.9)

Eq. (2.9) will be derived in Chapter 4, Eq. (4.21). H is the applied field, χm the measuredsusceptibility of the magnetic bead, and w = w/2z0 is the width of the sensor, w, nor-malized to the distance between sensor and bead center, z0. In this case, w = 1, becausew = 2 µm, and z0 = R = 1 µm, where R is the radius of the bead.

The sensor signals are given by Eq. (2.5) (planar Hall sensor) and Eq. (2.4) (spin-valvesensor). Table 2.2 summarizes the sensor response to a magnetic bead in the two cases ofsensors. The parameters for spin-valve sensors are from Ref. [26], parameters for planarHall sensors will be given in Chapter 6.

The combination of signal (Table 2.2) and noise (Table 2.1 and Fig. 2.4) is summarizedin Table 2.3 for selected frequencies. Clearly the planar Hall sensor outperforms the spin-valve sensor in the low frequency regime, where S/N is higher for the planar Hall sensor.However for large frequencies, when the spin-valve sensor is dominated by the white noiselevel, S/N is four times higher for the spin-valve sensor than for the planar Hall sensor.Their S/N match at f ≈ 1 kHz.


1 10 100 1000

0.1

1

10

100

PH

SV

PH

Noi

se d

ensi

ty (n

V/H

z1/2 )

Frequency (Hz)

White noise (PH) 1/f noise (PH) White noise (SV) 1/f noise (SV)

SV

Figure 2.4: Spectral noise for planar Hall sensor (PH) and spin-valve sensor (SV) ofcomparable geometry and materials.

AMR Ha (A m−1) H (A m−1) Hbead (A m−1) ∆Vbead (µV)PH 0.02 4000 1200 -34.64 -1.9SV 0.08 2400 1200 -34.64 12.5

Table 2.2: Signal from a single 2 µm magnetic bead at a planar Hall sensor (PH) anda spin-valve sensor (SV), both having 2µm × 2µm active sensing area, along with theparameters entering the calculations.

N30 ( nV√Hz

) S/N30 N1000 ( nV√Hz

) S/N1000 Nwhite ( nV√Hz

) S/Nwhite

PH 2.3 817 0.4 3161 0.45 4200SV 21.0 595 3.6 3376 0.7 18000

Table 2.3: Single bead signal-to-noise relations for planar Hall sensors compared to spin-valve sensors. N is the expected noise level. Selected frequencies are very low: f = 30 Hz,intermediate: f = 1 kHz, and very high: white noise dominated.

2.4. CHOICE OF SENSOR PRINCIPLE 19

2.4 Choice of sensor principle

In 2004, an investigation of the signal-to-noise for various magnetic biosensors was per-formed in the context of single bead detection by Freitas et al. [26]. The studied sensorswere designed for detection of a single 2 µm bead and had comparable active sensing areas.Planar Hall sensors with active sensing areas of 2.5µm × 2.5µm and spin-valve sensors,2µm× 3µm, were evaluated along with other magnetic field sensors, including GMR andHall sensors. The calculations were performed for the low frequency regime, f = 30 Hz,i.e. for spin-valve sensors dominated by the 1/f noise.

The results are listed in Table 2.4. The higher planar Hall S/N compared to my ownestimate is mainly due a lower effective anisotropy field, Ha = 2400 A m−1. The lowerS/N for spin-valve sensors comes from the 2 µm2 larger sensor area.

Sensor type GMR spin-valve Hall planar HallS/N 382 442 367 1453

Table 2.4: Signal-to-noise for competitive sensor technologies in magnetic biosensing, fromFreitas, 2004.

The most promising sensor in terms of bead detection limit is, according to this in-vestigation, the planar Hall sensor. However, going to high frequencies (f > 1 kHz, seeFig. 2.4) would reduce the 1/f noise of spin-valve sensors considerably, their S/N in thewhite noise regime would be approximately four times better than that of planar Hallsensors, see Table 2.3. Then again, when considering the design for a prototype deviceto be used in the doctor’s office, DC detection would probably be easier to handle onthe chip than high frequency lock-in detection. The planar Hall sensor offers the highestsignal-to-noise level with respect to DC bead detection.

Chapter 3

Theory of planar Hall sensors

This chapter describes the theory behind the planar Hall effect. The aim is to verify thatthe planar Hall sensor produces a linear signal as response to an applied magnetic fieldwithin a certain field range. Moreover, the hysteresis curves from an anisotropic materialare presented and related to the sensor signal. Finally, possible sources to thermal effectsare addressed.

The planar Hall effect stems from the anisotropic magnetoresistance effect of a magne-tized ferromagnetic material. Electrons travelling along or perpendicular to the magneti-zation experience different resistances. This gives rise to an anisotropic response. In orderto have a well defined starting point of the magnetization, an easy axis of magnetizationis introduced during metal deposition. The easy axis can be pinned by an antiferromag-netic material in order to obtain an easy direction of magnetization. The nickel sensorsconstituting our initial work on planar Hall sensors [2] experience uniaxial anisotropy,and the improved exchange biased permalloy sensors [4, 5] experience both uniaxial andunidirectional anisotropy.

3.1 The planar Hall effect

The planar Hall sensor is based on the planar Hall effect of ferromagnetic materials. Theresemblance to the ordinary Hall effect comes from the measuring geometry so the name issomewhat misleading. The origin of the planar Hall effect is from anisotropic scattering ofelectrons carrying current due to the magnetic moment of the lattice atoms. On the otherhand, the ordinary Hall effect is caused by the Lorentz force interaction between chargedparticles moving in a magnetic field. The planar Hall effect gives rise to an induced electricfield in the plane of the sensor when the anisotropic material is magnetized in the sameplane. Hence the name planar as opposed to the perpendicularly induced electric fieldacross a normal Hall element.

A drawing of the planar Hall sensor geometry is presented in Fig. 3.1. H is primarilyalong x. To begin with the magnetization vector lies in an arbitrary direction dependingon all coordinates (x, y, z).

21

22 CHAPTER 3. THEORY OF PLANAR HALL SENSORS

Figure 3.1: Geometry of the planar Hallsensor. The magnetization vector, M, hasan angle φ with respect to the x-axis. Mlies in the xy-plane, the current, I, is alongy, and the voltage, V , is measured alongx. The sensor cross has the width w andthickness t.

Figure 3.2: Decomposition of J into par-allel and perpendicular components withrespect to M. The angle between J andJ⊥ is φ.

The relevant physical relation is Ohm’s law,

E = ρJ (3.1)

Generally, the resistivity ρ is a tensor. In an anisotropic magnetoresistive (AMR) material,the following resistivities are encountered: ρH is the Hall resistivity, and ρ⊥ and ρ‖ are theresistivities when the magnetization vector is perpendicular and parallel to the currentdensity, respectively. For isotropic materials ρ⊥ and ρ‖ are equal but for ferromagneticmaterials, e.g. Ni, Fe, and Co, ρ‖ > ρ⊥ due to the AMR effect [6]. A unit vector is definedin the magnetization direction M, see Fig. 3.2, and the current density can be written interms of its parallel and perpendicular components,

J = J‖M + J⊥M⊥ (3.2)

J = J‖M + J− J‖M (3.3)

E is found from Eq. (3.1)

E = ρ‖J‖M + ρ⊥(J− J‖M) + ρHM× J (3.4)

By definition (Fig. 3.2), J‖ = J · M, and Eq. (3.4) can be written as [6]

E = M(J · M)[ρ‖ − ρ⊥] + ρ⊥J + ρHM× J (3.5)

The current density is chosen along the positive y-direction (see Fig. 3.1) such thatJ = Jyy. Jy = I/wt. I is the total current through the sensor, and w is the width

3.1. THE PLANAR HALL EFFECT 23

and t the thickness of the sensor. Particularly interesting to the sensor problem is Ex

which generates the voltage output of the sensor. In spherical coordinates (φ is defined inFig. 3.1, and θ is the polar angle measured from the z-axis) the magnetization unit vectoris

M =

sin θ cosφsin θ sinφ

cos θ

Ex can then be extracted from Eq. (3.5)

Ex = sin2 θ(ρ‖ − ρ⊥)Jy sinφ cosφ− ρH cos θMJy (3.6)

In the planar Hall sensor (Fig. 3.1) M is restricted to the xy-plane, and Eq. (3.6) reducesto

Ex = (ρ‖ − ρ⊥)Jy sinφ cosφ (3.7)

In terms of planar Hall voltage V = Ex ·w and total current, I = Jy ·wt, Eq. (3.7) becomes

V =I(ρ‖ − ρ⊥) sin φ cosφ

t=

I(ρ‖ − ρ⊥) sin 2φ2t

(3.8)

Now Eq. (3.8) can be rewritten in terms of the applied field and the anisotropic magne-toresistance coefficient (AMR ≡ ρ‖−ρ⊥

ρav), where ρav is the average resistivity of the film.

Defining x = cosφ, gives the following expression for the voltage drop across the width ofthe sensor

V = IAMR · ρav

tx√

1− x2. (3.9)

3.1.1 Magnetic energy and expected field response

Uniaxial anisotropy

It is assumed that H is essentially in the x-direction of Fig. 3.1 because the appliedexternal field is along x. The easy axis is along the y-direction, and the magnetization liesin the xy-plane at an angle φ with respect to the x-axis. For a ferromagnetic layer withuniaxial symmetry, using the first order approximation to the magnetic anisotropy givesthe following energy density of the film

E(φ) ≈ −µ0M ·H +µ0

2MHK cos2 φ (3.10)

In which the the anisotropy field is defined

HK ≡ 2Ku

µ0M(3.11)

The total magnetic field intensity (H) contains contributions from the applied field, thecurrent through the sensor, the field from beads, and possibly the fringing field from thesensor itself. Ku is the uniaxial anisotropy constant of the layer.


The z-component of the film’s magnetization is limited due to demagnetization effects.Furthermore, if the y-component of the magnetic field is small compared to the uniax-ial anisotropy field, Hy ¿ HK , only the x-component, Hx, needs to be considered andEq. (3.10) reduces to

E(φ) = −µ0MHx cosφ +µ0

2MHK cos2 φ (3.12)

In thermodynamical equilibrium, the energy of the film attains its minimum value, andcosφ can be found by minimizing the energy for |Hx| ≤ HK .

cosφ =Hx

HK(3.13)

For small applied fields, cosφ ¿ 1, the first order approximation of Eq. (3.9) becomes

V = IAMR · ρav

tx√

1− x2 ≈ IAMR · ρav

t

Hx

HK(3.14)

An expression for the sensitivity, S0 which is independent of applied parameters, of thesensor can be found

S0 ≡ Vx

HxI=

AMR · ρav

t ·HK(3.15)

Fig. 3.3 shows the magnetic field dependence of the sensor with uniaxial anisotropy. Thedotted line has the slope given in Eq. (3.15), which is accurate to within 2 % in the interval−0.20 < Hx/HK < 0.20.

Uniaxial and unidirectional anisotropy

The ferromagnetic layer can be pinned with an adjacent antiferromagnetic layer [27]. Thecontribution to the magnetic energy density can be regarded as an additional energy withunidirectional properties, i.e. the exchange energy. If an exchange field, HE , is added toEq. (3.12), the first order magnetic energy density of the film is

E(φ) ≈ −µ0MHx cosφ + 1/2µ0MHK cos2 φ− µ0MHE sinφ (3.16)

Minimization of Eq. (3.16) for small applied fields, i.e. angles close to φ = 90 o andsinφ ≈ 1, gives the cosφ expression

cosφ ≈ Hx

HK + HE(3.17)

HK is the anisotropy field, and HE is the exchange coupling field.For small applied fields the first order approximation of Eq. (3.9) becomes

V = IAMR · ρav

tx√

1− x2 ≈ IAMR · ρav

t

Hx

HK + HE(3.18)

And the sensitivity of the sensor becomes

S0 ≡ Vx

HxI=

AMR · ρav

t(HK + HE)(3.19)

3.1. THE PLANAR HALL EFFECT 25

-1.0 -0.8 -0.6 -0.4 -0.2 0.0 0.2 0.4 0.6 0.8 1.0-1.0

-0.5

0.0

0.5

1.0 Uniaxial

V/V

max

H/HK

Exact Linear

Figure 3.3: Expected sensor response as a function of applied magnetic field, H/HK . Thedotted line shows the linear approximation to Eq. (3.9).

-2.0 -1.5 -1.0 -0.5 0.0 0.5 1.0 1.5 2.0-1.0

-0.8

-0.6

-0.4

-0.2

0.0

0.2

0.4

0.6

0.8

1.0

HE/H

K=20

Unidirectional V/V

max

H/(HK+H

E)

Exact Linear

HE/H

K=0

Figure 3.4: Expected sensor response as a function of applied magnetic field, H/(HK+HE).The dotted line shows the linear approximation to Eq. (3.9). Also shown in the graph isthe HE = 0 special case of uniaxial anisotropy (grey line).


0.1 0.3 0.5 0.7 0.9 1.1 1.3 1.5-1.2

-0.9

-0.6

-0.3

0.0

0.3

0.6

0.9

1.2H

y=-H

K

Hy=-0.5H

K

Hy=0

Uniaxial energy (EA)

E(

)/2K

u

HK=0

Figure 3.5: Normalized film energy, E(φ)/2Ku, as a function of φ for different values ofHy/HK .

Eq. (3.9) is plotted in Fig. 3.4 with x defined in Eq. (3.16) for exchange biased filmanisotropy. The dotted line has the slope given in Eq. (3.19), which is accurate within 2% in the interval −0.23 < Hx/(HK + HE) < 0.23.

3.2 Hysteresis curves

The energy expression of Eq. (3.10) can be used to find the behavior of the film’s magne-tization in an applied magnetic field along the x-direction (hard axis) and the y-direction(easy axis). The magnetization vector depicted in Fig. 3.1 is

M = MS

(cosφsinφ

)(3.20)

where MS is the saturation magnetization of the film. For fields applied along the hardaxis of magnetization, this means that the x-component of the magnetization is

Mx = MS

HxHK

uniaxial |Hx| ≤ HK ,Hx

HK+HEunidirectional, small φ, |Hx| . 0.2(HK + HE).

(3.21)

3.2. HYSTERESIS CURVES 27

Easy axis hysteresis, uniaxial anisotropy

Considering fields applied along the easy axis of magnetization, the energy in uniaxialanisotropy is

E(φ) = µ0M(−Hy sinφ + 1/2HK cos2 φ) (3.22)

Eq. (3.22) is shown in Fig. 3.5 for different values of Hy/HK .For large positive fields, Hy À HK , the energy minimum is at sinφ = 1 ⇒ φ = π/2

(Fig. 3.5, HK = 0). Decreasing Hy destabilizes the minimum, and switching occurs wherethe second derivative disappears at Hy = −HK ,

d2E(φ)dφ2

|φ=π/2 = 0 ⇒

Hy

HK=

cos 2φ

sinφ|φ=π/2 (3.23)

Moving backwards, switching occurs at Hy = +HK ,

Hy

HK=

cos 2φ

sinφ|φ=−π/2 (3.24)

The resulting magnetization versus applied field curves are shown in Fig. 3.6 for appliedfield along the hard axis (HA) or along the easy axis (EA).

Easy axis hysteresis, exchange biased anisotropy

For the exchange biased film, with the field applied along the easy axis, the energy is

E(φ) = µ0M(−(Hy + HE) sin φ +12HK cos2 φ) (3.25)

which is identical to Eq. (3.22) exchanging Hy with Hy + HE .Switching of magnetization alignment from along +y to along −y occurs at

−(Hy + HE)/HK = 1 ⇒ Hy = −HK −HE

and switching from along −y to along +y occurs at

−(Hy + HE)/HK = −1 ⇒ Hy = HK −HE

The resulting magnetization versus applied field curves are shown in Fig. 3.7 both for theapplied field along the hard axis (HA) and along the easy axis (EA). The exchange fieldis visualized as a shift towards negative field values in the EA hysteresis loop.

Measurements of the hysteresis loops can be performed using, e.g., a vibrating samplemagnetometer (VSM). The hysteresis loop with the field applied along the easy axis ofmagnetization can thus be used to determine HK and HE , which can be used to predictthe sensor sensitivity using Eq. (3.19).


-2.0 -1.5 -1.0 -0.5 0.0 0.5 1.0 1.5 2.0

-1.0

-0.5

0.0

0.5

1.0

Uniaxial hysteresis (HA)

M/M

S

Hy/H

K

-2.0 -1.5 -1.0 -0.5 0.0 0.5 1.0 1.5 2.0

-1.0

-0.5

0.0

0.5

1.0

Uniaxial hysteresis (EA)

M/M

S

Hy/H

K

Figure 3.6: Hysteresis loops for uniaxial anisotropy. Left: applied field along the hardaxis. Right: applied field along the easy axis.

-2.0 -1.5 -1.0 -0.5 0.0 0.5 1.0 1.5 2.0

-1.0

-0.5

0.0

0.5

1.0

Unidirectional hysteresis (HA)

M/M

S

Hy/(H

K+H

E)

-1.0

-0.5

0.0

0.5

1.0

0

HK

Unidirectional hysteresis (EA)

M/M

S

Hy

-HE

Figure 3.7: Hysteresis loops for unidirectional anisotropy of an exchange biased film. Left:applied field along the hard axis. Right: applied field along the easy axis.

3.3 Effects of temperature

Investigation shows that the planar Hall sensor is temperature dependent. The tem-perature dependence is approximately linear in the temperature range of interest, andsubtracting the temperature drift from the signal should be fairly straight forward.

From Eq. (3.9) you will remember that five parameters enter the expression for theplanar Hall voltage: I, t, sinφ, AMR, ρav. A high precision current source is used tocontrol the current, and a thermal effect from the current can thus be ruled out. Thethickness of the ferromagnetic material will also remain constant, and sinφ, which isdetermined by the external field, will fluctuate very little at the temperatures of interest,i.e. those close to room temperature. This leaves us with the AMR ratio and the averageresistivity as possible contributors to a temperature dependence.

3.3. EFFECTS OF TEMPERATURE 29

AMR varies approximately 0.1 % oC−1 according to the authors of Ref. [9, 16]. Thiswas confirmed by personal communication with the late Prof. A. K. Menon. This variationis not enough to explain the temperature variation observed in the planar Hall sensors.

According to Bozorth [28] the resistivity of pure nickel changes considerably withtemperature. In Figure 16-24 he quotes data reported by Mott [29] where the resistivity(normalized to the value at the Curie point) changes approximately from 0.21 to 0.38 whenthe temperature changes from 0 oC to 100 oC. The curve is concave, not linear, but in asmall temperature window as the one the planar Hall sensor will experience, linearity is avery close approximation. A conversion of this sensitivity into a percentage value yields0.57 % oC−1.

The paper referred to, Ref. [29], proposes a model, where the resistivity is directlyproportional to temperature times a correction term, which depends on η, the ratio ofactual magnetization to saturation magnetization at T = 0 K

ρ = const.T

MΘ2

(1

(1− η)1/3 + α+

1(1 + η)1/3 + α

)−1

Θ is the Debye temperature, and α is a measure of transition probability between electronicorbitals s and d, at room temperature α ≈ 0.25. This correction term, in parentheses, ispresent but slowly varying at the temperature range of interest for the planar Hall sensor.Hence the contribution can be regarded as linear in planar Hall sensor applications. Itshould be noted that Mott obtains very close agreement with experiments when using thismodel.

The temperature dependence of the planar Hall effect in thin films was studied by Kyin 1968 [30]. Ky finds that the planar Hall effect increases with increasing temperature. Atemperature drift of approximately 0.2-0.25 % oC−1 is reported for 22 nm thick Ni filmsin the temperature range 200 K to 300 K.

An earlier experimental study by Shirakawa [31] finds that ∆Re drops approximately0.22 % oC−1 for the permalloy composition of bulk nickel-iron alloys. However, the findingsof Ky [30] for bulk nickel also show a decrease in the planar Hall effect with increasingtemperature, whereas the tendency is increasing for thin films. The parameters of thinfilms clearly differ from their bulk values.

For thin permalloy films, Montaigne et al. finds a temperature dependence in the orderof 0.3 % oC−1 for the isotropic part of the resistance, Re, [1].

Chapter 4

Theoretical sensor signal frommagnetic beads

This chapter presents a study of the single bead field with respect to sensor size andbead position both laterally and vertically above the sensor surface. This is followed bya summary of the theoretical detection limits based on the experimentally verified noiselevel. The aim is to verify the sensor’s capability of detecting just a single bead basedon the theoretical S/N. Hereafter, a study of a monolayer of non-interacting beads ispresented for the sensors used for influenza immunoassay detection. An estimate of thesignal magnitude produced by a monolayer of beads is obtained. Then a discussion ofdesigning the SU-8 layer optimally for bead detection experiments follows. What designis the best for bead detection experiments? Given the choice of SU-8 design, what signalcan be expected from a monolayer of beads?

For the medical diagnostic chips, the aim is simply a binary answer to the questionwhether beads are present or not. For the ensuing applications such as tracking cells, theaim is a truly quantitative measure of the number of beads in the sample. This chapteranalyzes the signal produced by a single magnetic bead placed above a general magneticsensor. In order to facilitate the computation and the interpretation of the results aquadratic sensor area is considered. It can readily be generalized to other geometries ofactive sensing area.

The position of the bead with respect to the sensor edges heavily influences the signalproduced. Centered above the sensor, the signal is negative, just before the sensor edgethe negative signal is enhanced, and passing the edge the bead gives a positive signal.This finding proposes a chip design that takes into account this change in signal whencrossing the edge. The signal from a monolayer of beads can be enhanced considerably byremoving the positively contributing beads.

4.1 Magnetic beads

This section describes the types of magnetic beads used for bead detection experiments.Their physical properties are described with the magnetic properties in focus. Included is

31

32 CHAPTER 4. THEORETICAL SENSOR SIGNAL FROM MAGNETIC BEADS

an estimate of when the assumption of non-interacting beads holds for the studied beadproducts.

Magnetic beads come in a large number of varieties and qualities. Those chosen forthis project are commercially available beads that exhibit nearly uniform properties: Dyn-abeads M-280 (www.dynal.no), Micromer-M and Nanomag-D (www.micromod.de) areclose to spherical and have nearly uniform magnetic properties. They are polystyrenespheres with nanoparticles of iron oxide embedded in the matrix. The surface of thebeads can be functionalized with various biochemical species, nucleic acids (DNA andRNA), proteins (specific antibodies), or even cells, which are all attached to the surface ofthe magnetic bead. Another option is to put magnetic beads inside cells by endocytosis.When the acquired species are related to the presence of the bead, attached to the surfaceof the bead, or filled up with beads, the presence and the whereabouts of the species canbe revealed by tracking the bead.

4.1.1 Magnetic characterization of beads

Magnetic characterization of the beads is performed using a Lakeshore Model 7407 vibrat-ing sample magnetometer (VSM). Fig. 4.1 shows the curve obtained for 18.95 mg of 250nm beads, Nanomag-D. The size refers to the average diameter of the beads. Analysis ofthe low field slope with compensation for the diamagnetic contribution by the containeryields the measured magnetic susceptibility, χm = 3.2± 0.1 (SI). The uncertainty on χm

is estimated from the slope and assumes zero uncertainty on the bead density. Due todemagnetization effects, χm > 3 is not possible if the beads are completely spherical.

The magnetic properties of various commercially available beads investigated with theVSM, and the physical properties provided by the supplier, are summarized in Table 4.1.

4.1.2 Bead-bead interaction

The assumption that the beads are not interacting is used throughout the study of thesignal produced by a monolayer of beads. This assumption is valid, when the field producedby a bead is small compared to the external field, i.e. H À Hbead, which for a dipole fieldmeans [32]

H À MR3

3r3=

χmHR3

3r3

r is the distance from one bead center to the next, R the bead radius, and M is themagnetization. For a linear magnetic material M = χmH, where χm is the measuredsusceptibility. Rearranging, we get

( r

R

)3À χm

3(4.1)

For Nanomag-D beads, χm ≈ 3, we get that the assumption of non-interacting beads holdswhen

r & R

The validity of the non-interacting beads approximation is summarized in Table 4.1 forthe studied beads.

4.1. MAGNETIC BEADS 33

-1500 -1000 -500 0 500 1000 1500

-1.5x105

-1.0x105

-5.0x104

0.0

5.0x104

1.0x105

1.5x105

-5 0 5

-3x104

-2x104

-1x104

0

1x104

2x104

3x104

= 3.2 (0.1)

NanomagD 250 nm

Mag

netization

(A m

-1)

Applied field (kA m-1)

Figure 4.1: Vibrating sample magnetometry (VSM) measurements of 18.95 mg Nanomag-D beads, average diameter 250 nm. The insert shows the low-field response of the magneticbeads, where the slope is an estimate of the susceptibility of the beads. A remanentmagnetization is present in the sample.

Bead Diameter Concentration Density χmrR &

Dynabeads M-280 2.8 µm 10 mg ml−1 1.3 g cm−3 0.13 (SI)* 0.35Micromer-M 2 µm 25 mg ml−1 1.4 g cm−3 0.3(0.1) (SI) 0.46Nanomag-D 250 nm 10 mg ml−1 2.5 g cm−3 3.2(0.1) (SI) 1.02

Nanomag-D-spio 50 nm 10 mg ml−1 1.4 g cm−3 1.4(0.1) (SI) 0.78

Table 4.1: Physical properties of magnetic beads given by the supplier or measured ona VSM. The errors are determined from the slope assuming no error on the density. χm

is the measured magnetic susceptibility. (*) χm of the Dynabeads is estimated from theinformation given by the supplier (www.invitrogen.com). r

R is the bead-bead distancedivided by the radius, tabulated are the values to which extent the non-interacting beadapproach holds, Eq. (4.1).


4.2 Sensor and bead

After looking at their physical properties, the beads are considered with respect to amagnetic sensor. The planar Hall sensor, spin-valve sensor, and GMR sensor are magneticfield sensors sensitive to fields in the plane of the sensor. A linear magnetic field sensor issensitive to the average field above the active sensing area. Therefore the mean value ofthe magnetic field produced by the bead is calculated for certain sensor geometries.

An applied external field magnetizes the bead. The field strength from the bead justbelow the bead opposes the external field. The field just below the bead represents amaximum value that can be detected by the sensor if it were possible to construct aninfinitely small sensor. The planar Hall sensor uses all its active area for detection, whichmeans that the field felt by the sensor is an average of the field strength at all locationsof the sensitive area. This average is calculated by a weighted integral.

4.3 Analytical study of single bead signal

In this section the average bead field, 〈Hx〉, is calculated. Assuming uniform magneticproperties and spherical geometry, the field produced by a magnetic bead is the dipolefield. The linear sensor measures the x-component of the field averaged over the sensorarea. Therefore the measured field is estimated as the average of Hx, where the averageis produced over the sensor area. Two sensor geometries, circular and quadratic, areevaluated.

The magnetic field strength produced by a single bead (assuming it is spherical andcan be regarded as uniformly magnetized) magnetized by an external field, H0x, is thewell-known dipole field derived in Appendix A.

H =M

3R3

r3(3(M · r)r − M) (4.2)

M is the magnetization, and M a unit vector in the direction of magnetization, i.e. alongx. R is the bead radius, and r the distance from the center of the bead to the observationpoint. The geometry is sketched in Fig. 4.2, omitting M for clarity. In the following, x andr are unit vectors in the cartesian and spherical coordinate systems, respectively. ρ andφ are unit vectors in an in-plane polar coordinate system, with φ defined with respect tothe x-axis. The spherical coordinate system consists of r, θ, and φ, where the azimuthalangle φ is identical to φ in the polar coordinate system, and θ is the polar angle definedwith respect to the z-axis.

Due to demagnetization effects, which are opposing magnetization components per-pendicular to the sensor plane, only in-plane field components are important. The y-component of the field produced by the bead is small compared to the effective anisotropyof the film, and thus only the x-component needs to be considered. The x-componentof the field produced by the bead is sensed by the planar Hall sensor. Rememberingx · r = sin θ cosφ, and x · M = 1, the x-component of the field can be extracted from

4.3. ANALYTICAL STUDY OF SINGLE BEAD SIGNAL 35

Figure 4.2: Schematic drawing of a bead of radius R placed above a magnetic sensor. ris the distance from the center of the bead to the observation point, z0 is the verticaldistance from the center of the bead to the sensor, and ρ is the distance in the sensorplane from the center of the bead to the observation point.

Eq. (4.2).

Hx = H · x =M

3R3

r3(3 sin2 θ cos2 φ− 1) (4.3)

In the case of maximum produced field at a right angle to the bead magnetization r = z0,where z0 is the normal distance along z, see Fig. 4.2. The maximum value found fromEq. (4.3) is

Hmax = −M

3R3

z30

(4.4)

If, additionally, the field is measured just below the bead, at z0 = R, Hmax reduces to−M/3.

The trigonometric sine function in Eq. (4.3) can be rewritten in terms of the distancesinvolved as sin θ = ρ

r , where ρ is the distance in the xy-plane identical to the radius in thepolar coordinate system. With these definitions Eq. (4.3) becomes

Hx =MR3

3(3ρ2 cos2 φ− r2)

r5(4.5)

Using Pythagoras’ relation eliminates the last mixed variable: r =√

ρ2 + z20

Hx =MR3

33ρ2 cos2 φ− (ρ2 + z2

0)(ρ2 + z2

0)5/2(4.6)

To obtain an estimate of the field strength influencing the sensor, Hx is integrated over ageneral sensor area, A, below the bead, the integral is evaluated at a distance z0 ≥ R.

〈Hx〉 =1A

∫HxdA (4.7)

where A is the sensor area and dA the differential area element.


4.3.1 Circular sensor geometry

For a circular sensor cross section of radius ρ0 with the bead placed above the center,the integral turns out to be fairly straightforward. In polar coordinates dA = ρdρdφ, andA = πρ2

0.

〈Hx〉 =1

πρ20

∫ 2π

0

∫ ρ0

0Hxρdρdφ (4.8)

〈Hx〉 =1

πρ20

∫ 2π

0dφ

∫ ρ0

0ρdρ

MR3

33ρ2 cos2 φ− (ρ2 + z2

0)(ρ2 + z2

0)5/2(4.9)

First the φ-integral is evaluated, in which the integral of cos2 φ is π, and the integral of 1is 2π. Eq. (4.9) thus reduces to

〈Hx〉 =MR3

3ρ20

∫ ρ0

0ρdρ

(3ρ2 − 2(ρ2 + z20))

(ρ2 + z20)5/2

(4.10)

〈Hx〉 =MR3

3ρ20

∫ ρ0

0dρ

(ρ3 − 2ρz20)

(ρ2 + z20)5/2

(4.11)

Introduction of the dimensionless variables ρ = ρ/z0 and R = R/z0 yields the dimension-less integral

〈Hx〉 =M

3R3

ρ02

∫ fρ0

0dρ

(ρ3 − 2ρ)(ρ2 + 1)5/2

(4.12)

This, normalized to |Hmax| = R3M/3, gives

〈Hx〉 =1

ρ02

∫ fρ0

0dρ

(ρ3 − 2ρ)(ρ2 + 1)5/2

(4.13)

Eq. (4.13) can be solved analytically

〈Hx〉 = − 1(1 + ρ0

2)3/2(4.14)

Eq. (4.14) shows that the average field produced by a magnetic bead is indeed negative. Italso shows that for large sensor areas or very small bead diameters, where edge effects areunimportant, the signal from the bead scales inversely to sensor size cubed, 〈Hx〉 ∝ ρ−3

0 .When considering the z0 dependence, the normalization of R should be kept in mind, andwe get a scaling which is inversely proportional to separation distance cubed, 〈Hx〉 ∝ z−3

0 ,for large separations compared to the sensor size.

The result of Eq. (4.14) is plotted in Fig. 4.3 as a function of sensor size, ρ0.

4.3.2 Quadratic sensor geometry

The bead is placed in the center of the sensor at (x, y) = (0, 0). The sensor has the areaA = w2. In cartesian coordinates Eq. (4.6) becomes

Hx =MR3

32x2 − y2 − z2

0

(x2 + y2 + z20)5/2

(4.15)

4.3. ANALYTICAL STUDY OF SINGLE BEAD SIGNAL 37

0.0 0.5 1.0 1.5 2.0 2.5 3.0-1.0

-0.8

-0.6

-0.4

-0.2

0.0 Circular sensor geometry

Hx/H

max

0z

0

Figure 4.3: Scaling of 〈Hx〉 for a single bead at the center of a circular sensor with sensorsize ρ0 = ρ0/z0, Eq. (4.14).

0.0 0.5 1.0 1.5 2.0 2.5 3.0-1.0

-0.8

-0.6

-0.4

-0.2

0.0

Quadratic sensor geometry

Hx/H

max

w/2z0

Figure 4.4: Scaling of 〈Hx〉 for a single bead at the center of a quadratic sensor with sensorsize w = w/2z0, Eq. (4.19).


And Eq. (4.7), the average of Eq. (4.15), is

〈Hx(w, z0)〉 =MR3

3w2

∫ w/2

−w/2dx

∫ w/2

−w/2dy

2x2 − y2 − z20

(x2 + y2 + z20)5/2

(4.16)

Again the introduction of dimensionless variables (x = x/z0, y = y/z0, w = w/2z0, and〈Hx〉 = 〈Hx〉/Hmax, where Hmax = R3M/3) simplifies the expression to

〈Hx(w)〉 =1

4w2

∫ ew

− ewdx

∫ ew

− ewdy

2x2 − y2 − 1(x2 + y2 + 1)5/2

(4.17)

=1

w2

∫ ew

0dx

∫ ew

0dy

2x2 − y2 − 1(x2 + y2 + 1)5/2

(4.18)

The result of the integration is

〈Hx(w)〉 = − 1(1 + w2)(1 + 2w2)1/2

(4.19)

〈Hx〉 as a function of w is plotted in Fig. 4.4. For large values of w, 〈Hx〉 approaches zero.Eq. (4.19) proves the initial intuition that the signal from a magnetic bead centered

above a magnetic sensor is negative. According to this equation the average field fromthe bead scales as the inverse cube of the sensor size, 〈Hx〉 ∝ w−3, for large sensor sizescompared to the diameter of the bead. Fig. 4.4 shows that the signal from the beaddecreases rapidly as a function of sensor size. The ideal geometry of the sensor for singlebead detection would thus be infinitely small, or small compared to the bead diameter.

4.3.3 Scaling with sensor size

Comparing Fig. 4.3 and Fig. 4.4 one can observe that the two sensor geometries behavesimilarly. It is only for sensor sizes near z0 the two figures differ. For large values, Table4.2 summarizes the scaling of ρ0, w, and z0.

Parameter Scalingρ0 〈Hx〉 ∝ ρ−3

0

w 〈Hx〉 ∝ w−3

z0 〈Hx〉 ∝ z−30

Table 4.2: Scaling of the average bead field, 〈Hx〉, with sensor size ρ0 (circular sensor) andw (quadratic sensor) and sensor-bead separation z0 for a bead centered on the sensor.

4.4 Single bead detection limits for 20µm× 20µm sensors

In this section an investigation of the signal-to-noise relation for various commerciallyavailable beads is performed. First the actual sensor properties are described. Then the

4.4. SINGLE BEAD DETECTION LIMITS FOR 20µM× 20µM SENSORS 39

0.0 0.5 1.0 1.5 2.0

0.95

0.96

0.97

0.98

0.99

1.00

Effect of bead sensor separation for Nanomag-D beads (250 nm)N

ormalized

bea

d sign

al V

z/V0.5

Bead-sensor distance (µm)

Figure 4.5: Effect of increasing the separation of the sensor and the bead. The beadis placed at the center of a quadratic sensor with w = 20 µm, and the average field iscalculated from Eq. (4.19). The bead signal is normalized to the value at z0 = R, wherethe bead is placed directly on top of the sensor.

single bead S/N is calculated for three commercial bead products. Finally, the sensor sizeis varied. The smaller the sensor, the higher the single bead S/N is found.

Eq. (4.19) gives the average field from a single bead of arbitrary size centered abovea general magnetic sensor. Eq. (4.4) describes the properties of a specific bead, recallingthat M = χmH. The χm-values for the beads are in Table 4.1. If a specific sensor isconsidered, w = w/2z0 defines the geometry of the sensor with respect to the lateral beadposition, and S0 is the sensor sensitivity. As an example, the fabricated planar Hall sensorsare 20µm× 20µm, i.e. w = 20 µm.

The signal produced by a bead, Vbead, is thus

Vbead = 〈Hx〉S0 (4.20)

Vbead = − S0R3χmH

3(1 + w2)(1 + 2w2)1/2(4.21)

Fig. 4.5 visualizes the effect of increasing the separation of the sensor and a single Nanomag-D bead (250 nm in diameter). In this figure, the bead signal is normalized to its valueat z0 = R, and the bead-sensor distance is the increase in distance from z0 = R. SU-8 is


needed on the surface for binding chemistry, and a layer yielding z0 = 0.5 µm is chosen asa reasonable tradeoff between single bead signal and the ability to do surface chemistry.

The experimental conditions are B = 0.5 mT and I = 1 mA giving S0 = 44 µV mT−1

(Chapter 6). Under these conditions, a Nanomag-D bead placed at the sensor center andat z0 = 0.5 µm will produce a signal of

Bbead = −7.3 · 10−10 T = −0.73 nT

Hbead = −5.8 · 10−4 A m−1

Vbead = −3.2 · 10−11 V = −0.032 nV

The experimental noise level for magnetic immunoassaying is ∆V ≈ 1 nV, which is thestandard error on the average of 100 samples (Chapter 7). Physical and magnetic proper-ties can be found in Table 4.1.

The single bead signal-to-noise relations and corresponding bead detection limits, re-alized as the number of beads needed to give S/N = 2, are presented in Table 4.3 and inFig. 4.6. The single bead signal is given by Eq. (4.21). Table 4.3 describes what can beexpected of the actual planar Hall sensors used in chapters 6, 7, and 8. The monolayerS/N assumes non-interacting beads and that 〈Hx〉 is invariant with position. The lastassumption is inaccurate, but will be addressed in detail in later sections. Note that S/Nproduced by a monolayer of Nanomag-D beads exceeds that of Micromer-M in spite oftheir much lower single bead S/N.

Fig. 4.6 shows S/N and detection limits for the three commercial bead products asa function of sensor size. With areas smaller than 1 µm2 it is theoretically possibleto approach the detection of a single 50 nm magnetic label. Remarkably, this result isobtained optimizing only the sensor size, though the analysis assumes that a current ofIAC = 1 mA can be applied to all sensors including the smallest. 10µm × 10µm sensorscan sustain 10 mA DC [4], indicating that a 1µm× 1µm sensor is likely to sustain 1 mADC.

4.5 Quadratic sensor and any bead position

This section investigates how the bead’s position influences the average field it produces.The bead is placed at (x′, y′), different from the integration variables (x, y), and scannedacross the sensor surface. First a small sensor is considered, then a 20µm× 20µm sensor.

For a bead situated at an arbitrary position above the quadratic sensor, the integralsare solved numerically. Mathematica code is included in Appendix C.

4.5. QUADRATIC SENSOR AND ANY BEAD POSITION 41

Bead Diameter S/N Detection limit S/N per monolayerMicromer-M 2 µm 1.55 1-2 155Nanomag-D 250 nm 0.032 62 206

Nanomag-D-spio 50 nm 1.1 · 10−4 17700 18

Table 4.3: Single bead S/N and detection limits for a 20µm × 20µm planar Hall sensor.Experimental parameters for noise level of 1 nV, B = 0.5 mT, I = 1 mA, and χm in Table4.1. The bead signal is obtained from Eq. (4.21). The detection limit is the number ofbeads needed to give S/N = 2. The monolayer S/N assumes non-interacting beads andthat the average field does not change with bead position within the sensor area.

1 10

1E-4

1E-3

0.01

0.1

1

10

100

1000

50 nm

250 nm

Signal to noise

S/N

Sensor size (µm)

2 µm

1 101E-4

1E-3

0.01

0.1

1

10

100

1000

10000

2 µm

250 nm

Bead detection limit

Det

ectio

n lim

it

Sensor size (µm)

50 nm

Figure 4.6: Single bead signal-to-noise relations and bead detection limit for increasingsensor sizes. The beads are placed at the center of a square sensor, and the sensor size refersto the edge length. The bead signal is obtained from Eq. (4.21). The bead detection limitis realized as the number of beads needed to give S/N = 2 for the specific experimentalconditions described in the text. The bead types are: Micromer-M (2 µm), Nanomag-D(250 nm), Nanomag-D-spio (50 nm). (Properties in Table 4.1.)


10

20

30

40

10

20

30

40

-0.2

-0.1

0

0.1

10

20

30

4010 20 30 40

10

20

30

40

Figure 4.7: Visualization of the numerical values of 〈Hx(x′, y′)〉 obtained from evaluationof Eq. (4.22). The x-y-axes show the matrix entry number. Left: Landscape plot for−2 < x′, y′ < 2. Right: Contour plot for −2 < x′, y′ < 2. White is the highest valueattained, black is the lowest. The sensor edges are placed x′ = ±1 and y′ = ±1, which inthe matrix corresponds to 10 and 30, respectively.

4.5.1 Small sensor compared to bead size

Choosing a sensor size of w = 1 places the sensor edges at (x′, y′) = ±1. (Normalizedcoordinates: x = x/z0, y = y/z0, w = w/2z0, and 〈Hx〉 = 〈Hx〉/Hmax, where Hmax =R3M/3.) The signal from the bead as a function of arbitrary bead position, x′ and y′, is,when rewriting Eq. (4.17),

〈Hx(x′, y′)〉 =14

∫ 1

−1dx

∫ 1

−1dy

2(x′ − x)2 − (y′ − y)2 − 1((x′ − x)2 + (y′ − y)2 + 1)5/2

(4.22)

The integral is evaluated for x′ ∈ [−2; 2] and y′ ∈ [−2; 2] in order to observe the effect ofcrossing the edges at x′ = ±1 and y′ = ±1. The numerical values are presented in Fig. 4.7.

From the figures it can be observed that a bead placed in the center of the sensorproduces the maximum |〈Hx〉|-value. From the list of integrals (data not shown), it canbe seen that a bead placed at certain positions (at the edges, x′ = ±1) the contributionchanges sign from negative to positive, and a bead placed here will thus decrease the total|〈Hx〉| from a layer of beads distributed evenly.

4.5.2 20µm× 20µm sensor

In order to analyze a specific sensor, the generalized form of Eq. (4.22) is considered

〈Hx(x′, y′)〉 =1

4w2

∫ ew

− ewdx

∫ ew

− ewdy

2(x′ − x)2 − (y′ − y)2 − 1((x′ − x)2 + (y′ − y)2 + 1)5/2

(4.23)


(Normalized coordinates: x = x/z0, y = y/z0, w = w/2z0, and 〈Hx〉 = 〈Hx〉/Hmax, whereHmax = R3M/3.) Int will be defined as the integral

Int(x′, y′) =∫ ew

− ewdx

∫ ew

− ewdy

2(x′ − x)2 − (y′ − y)2 − 1((x′ − x)2 + (y′ − y)2 + 1)5/2

(4.24)

The normalized field from a single bead is thus defined as

〈Hx(x′, y′)〉 =Int(x′, y′)

4w2(4.25)

The sensors of chapters 6, 7, and 8 are studied. The dimension of the sensors is 20µm ×20µm and they are covered by a thin layer of SU-8, used for the binding chemistry. TheSU-8 layer is constructed such that the bead’s center is placed a distance z0 = 0.5 µmabove the sensor. The beads detected in chapters 7 and 8 are the Nanomag-D beads,of diameter 250 nm. Using these specific dimensions the dimensionless parameters aredetermined as: w = 20 and R = 0.25.

Using the specific values for the dimensionless parameters, the integral is

Int(x′, y′) =∫ 20

−20dx

∫ 20

−20dy

2(x′ − x)2 − (y′ − y)2 − 1((x′ − x)2 + (y′ − y)2 + 1)5/2

(4.26)

and the average field from a Nanomag-D bead situated at (x′, y′)

〈Hx(x′, y′)〉 =R3M

3· 〈Hx(x′, y′)〉 =

R3M

3· Int(x′, y′)

4w2(4.27)

The integral, Int, is symmetric with respect to x′ and y′, such that Int(−x′) =Int(+x′) and Int(−y′) = Int(+y′). Therefore only the expression in the positive quadrantis evaluated: x′ ∈ 0, 40, and y′ ∈ 0, 40. Expecting the x′-value of the integral, Int,to give rise to the greatest variations, the integral along x′ from the center of the sensorpast the edge is considered. When the center of the bead reaches the edge of the sensorHx goes from contributing negatively to the integral to contributing positively. This signchange is expected to have a great impact on the value of Int. Some positive contributionsare removed from the integral, when the positive part of the bead field is placed outsidethe sensor.

Fig. 4.8 shows Int as a function of bead position starting from the center of the sensor,(x′, y′) = (0, 0), and continuing across the edge with varying bead position along x (blueline). To underline that the change almost solely depends on the x position of the bead,Int along y is also included on the graph (red line). The edge, positioned at x = 10 µm(or y = 10 µm), is marked with a horizontal dotted line.

The first part of the path along x is relatively flat, then two parts with peaks inopposite directions follow, and in the end again a relatively flat path. It is clear from thebehavior of Int along x that the absolute x-position of a bead with respect to the sensorhas a strong influence on the field felt by the sensor. If beads are only situated inside thesensor’s active area, their contribution to the applied field will be negative. If, however,beads have crossed the line and are positioned outside the sensor’s active area, they willgive a positive contribution to the field.


0 5 10 15 20

-1.0

-0.5

0.0

0.5

1.0

Along y

Integral along x, y

Int

Bead position (µm)

Along x

Figure 4.8: Normalized integral of magnetic field strength, Int, from a single bead withvarying position with respect to the center of the sensor. Along x (x′, 0), and along y(0, y′). The edge of the sensor is marked with a dotted line at x, y = 10 µm.

4.5.3 Monolayer of Nanomag-D beads on 20µm× 20µm sensors

When the sensors are used for measuring the layer coverage resulting from bead detectionexperiments, it is important to know how a specific coverage of beads influences the sensorsignal. The number of beads, which can be closely packed on a 20µm × 20µm sensor, is7390. The sum of contributions from each individual bead in a monolayer of beads isestimated as the trapezoidal sum Trp of the integrals Int (Eq. (4.26)). The sum assumesnon-interacting beads plus a statistical layer coverage, in which the beads are distributedevenly over the evaluated surface.

Trp =∑

i,j

0.25(Inti,j + Inti+1,j + Inti,j+1 + Inti+1,j+1) (4.28)

Bead-bead interactions are disregarded in this evaluation. It should be noted thatbead-bead interactions are expected to decrease the average field produced by a layer ofbeads in comparison to non-interacting beads. This is because the field produced by asingle bead opposes the applied field, and hence lowers the field felt by its neighbor.

Since the variation of the bead signal is strongest near the sensor edges with respectto x, the sensor area is divided into two edge parts, Ei and Eo (inside and outside, respec-tively), and four parts, where the integral varies more slowly, C, F1, F2, F3 (central andflat 1, 2, 3, respectively).


Choosing the area to be evaluated equal to that spanned by x and y in Fig. 4.8,yields x ∈ 0µm, 20µm and y ∈ 0µm, 20µm, which is identical to x′ ∈ 0, 40, andy′ ∈ 0, 40, as mentioned previously. This area is divided into six parts, listed in Table4.4 and shown in Fig. 4.9. In terms of the x′ value, the border between C and Ei is set atInt = −0.5 in Fig. 4.8 and equivalently is the border between Eo and F3 at Int = +0.5in Fig. 4.8. For the y′ value, the border is chosen at 2 µm past the edge facilitating theevaluation of the integrals in F1 and F2.

The field from a monolayer of beads covering one of the areas is calculated using thefollowing equation,

〈Hx〉 = 4 · 7390MR3

3 · 4w2

A

Atot

Trp

n(4.29)

where M is the bead’s magnetization, R = R/z0 the normalized bead radius, w = w/2z0

the normalized sensor width, A the area of the evaluated part, Atot the total area ofC+Ei+Eo+F (where F=F1+F2+F3), Trp the trapezoidal sum, and n the number of in-tegrals in the sum.

Normalized to the applied flux density (in mT), the normalized value β reflects theflux density produced by the beads in any applied field,

β ≡ µ0〈Hx〉B(mT)

= 4 · 7390χmR3

3 · 4w2

A

Atot

Trp

n(4.30)

as long as the beads’ susceptibility, χm, is linear with the applied field, i.e. for smallapplied fields.

The trapezoidal sum, Trp, of all Int’s in an area like C is evaluated numerically andpresented in Table 4.5. The intervals at which an integral is evaluated, ∆x′ and ∆y′, arelisted along with the number of Int’s in the sum, n. n can be related to the number ofbeads on the area by nbeads = n ·7390/4221 for a monolayer of beads. β is calculated fromEq. (4.30) for a monolayer of beads covering the area in question. Included is the otherthree quadrants giving identical contributions to the evaluated one. Marked in bold arethe expected values for a full coverage of beads at a distance of z0 = 0.5 µm (Fig. 4.10-1).This is the sum of all contributions positive as well as negative.

The average field from a monolayer of beads is negative. This is expected since thenegative contributions from the bead field should dominate the positive contributions.

The last two lines of Table 4.5 gives the signal from the beads, when only the sensor(C+Ei) or outside the sensor (Eo+F) is covered with beads. It is evident that not muchmisplacing of beads is required to flip the sign of the sum from negative to positive. Thetwo contributions have almost the same magnitude though with opposite signs. If mostof the beads from Ei were instead placed in Eo, the sign of the field from the beads wouldchange.


Position x′ y′ x′ y′

C Central area (0-8) µm (0-12) µm (0-16) (0-24)Ei Edge, inside (8-10) µm (0-12) µm (16-20) (0-24)Eo Edge, outside (10-11.7) µm (0-12) µm (20-23.4) (0-24)F1 Flat 1 (0-11.7) µm (12-20) µm (0-23.4) (24-40)F2 Flat 2 (11.7-20) µm (12-20) µm (23.4-40) (24-40)F3 Flat 3 (11.7-20) µm (0-12) µm (23.4-40) (0-24)

Table 4.4: Division of total area, Atot, of the positive quadrant of the sensor plus thearea in its immediate vicinity. Atot is divided in terms of the magnitude and sign of therespective contributions to the total signal from a monolayer of beads.

Figure 4.9: Area division of the evaluated trapezoidal sum. The red lines and charactersshow this division listed in Table 4.4. The coordinate system is placed at (x, y) = (0, 0)in the center of the sensor (the sensor border is indicated with a dotted black line). Thecontour plot shows the integral, Int, as a function of position of the bead in the criticalarea where the integral peaks. It is included in order to illustrate the position of theboundaries between positive, negative, and ”flat” areas.

4.6. DESIGN OF SU-8 LAYER 47

∆x′ ∆y′ n = nx · ny Trp β monolayerC 0.2 2 1053 -182.238 -0.0128Ei 0.2 2 273 -143.844 -0.0097Eo 0.2 2 234 +126.033 +0.0085F1 0.2 2 1062 -32.9518 -0.0022F2 0.2 2 756 +9.81117 +0.0007F3 0.2 2 1092 +140.949 +0.0099All 0.2 2 4221 -82.2856 -0.0057

C+Ei 0.2 2 1313 -326.127 -0.0225Eo+F 0.2 2 2908 243.841 +0.0168

Table 4.5: Trapezoidal sums, Trp, of the integrals, Int, and the average magnetic field,β = µ0〈Hx〉/B, Eq. (4.30), produced by a monolayer of beads covering the respectivearea. ∆x′ and ∆y′ are the interval lengths, and n is the number Int’s contributing tothe trapezoidal sum of integrals. n can be related to the number of beads on the area bynbeads = n · 7390/4221 (for a close packed monolayer - see text for details).

4.6 Design of SU-8 layer

This section discusses the design of the SU-8 layer used for binding chemistry. The con-straint of the design is that z0 = 0.5 µm on top of the sensor area. This distance is chosenas a reasonable compromise between single bead signal (Fig. 4.11) and the ability to dosurface chemistry.

Fig. 4.10 shows various possible designs of the SU-8 layer, The design denoted (1),where z0 = 0.5 µm throughout the entire chip, is used to obtain the results listed in Table4.5. The figure also shows two alternative designs of the SU-8 layer (2) and (3). (2)corresponds to the design of the sensors used for immunoassay experiments, and (3) is theprotective capping layer design, where the positive contributions from beads adjacent tothe sensor have been diminished using an SU-8 capping layer.

In the following, the average field produced by a monolayer of beads covering design(2) and (3) will be analyzed. The parameter z0 will be varied on the sensor area (C+Ei)or outside the sensor area (Eo+F). Hence, the integration will be performed in ordinarycoordinates as opposed to normalized.

The signal from a monolayer of beads covering an area is calculated using the followingequation.

β =µ0〈Hx〉B(mT)

= 4 · 7390χmR3

3w2

A

Atot

Trp′

n(4.31)

in which the trapezoidal sum, Trp′, sums the non-normalized integrals, Int′,

Int′(x′, y′) =∫ 10

−10dx

∫ 10

−10dy

2(x− x′)2 − (y − y′)2 − z20

((x− x′)2 + (y − y′)2 + z20)5/2

(4.32)

and the rest of the parameters have been defined in connection with Eq. (4.29) andEq. (4.30). Eq. (4.31) assumes non-interacting beads.


Figure 4.10: Design of the SU-8 capping layer. (1): 0.5 µm covering all. (2): 0.5 µm SU-8only covering the sensor area. (3): protective capping layer of zcap = 5 µm adjacent tothe sensor.

Design 1

Design 1 has been evaluated in Table 4.5. For a monolayer coverage β = −0.006 isobtained. Large negative values from the central area (C+Ei, β = −0.023) dominateslarge positive values from outside the sensor (Eo+F, β = +0.017). Therefore this designis sub-optimal in terms of bead detection.

Design 2

The best possible situation for bead detection would be only to cover the sensor area withbeads. This way the positive contribution from Eo+F would vanish.

The average field of such an arrangement of a monolayer of beads is calculated usingEq. (4.32) on area C and Ei. A numerical evaluation of the integrals is conducted. Theresults are shown in Fig. 4.11 for increasing sensor-bead distance, z0. The maximumobtained value for z0 = 0.125 µm is β = −0.030. The field produced by a monolayer ofbeads at z0 = 0.5 µm is β = −0.023.

The monolayer field resulting from the SU-8 design with z0 = 0.5 µm is thus reduced26 % with respect to the maximum obtainable value.

Design 3

Another approach could be to shield the positive contribution from beads outside thesensor area, i.e. lift area Eo+F to a distance, where the positive contribution is diminished.Fig. 4.10 (3) shows the design of the SU-8 layer for this approach. zcap is the cappinglayer covering Eo+F. The relation to z0 is zcap = z0 −R, where R is the bead radius.

zcap = 0 µm coincides with design (2), but this time beads adjacent to the sensor areincluded. Eq. (4.32) is evaluated on area C and Ei for z0 = 0.5 µm. On area Eo and F

4.6. DESIGN OF SU-8 LAYER 49

0 1 2 3 4 5

-0.030

-0.025

-0.020

-0.015

-0.010

-0.005Monolayer covering sensor area

z0 (µm)

Figure 4.11: Theoretically expected signal, β = µ0〈Hx〉/B, Eq. (4.31), from a monolayerof Nanomag-D beads covering only the sensor area as a function of distance, z0. Thedesign is shown in Fig. 4.10 (2).

0 1 2 3 4 5

-0.020

-0.015

-0.010

-0.005

0.000

0.005

Shielding of positive contributions by zcap

zcap

(µm)

Figure 4.12: Theoretically expected signal, β = µ0〈Hx〉/B, Eq. (4.31), from a monolayerof Nanomag-D beads with SU-8 capping layer outside the sensor, zcap. The design of zcap

is shown in Fig. 4.10 (3).


Eq. (4.32) is evaluated for z0 defined by zcap.Fig. 4.12 shows the average field dependence as a function of increasing SU-8 capping

layer. The average field at zcap = 0 µm is β = +0.003, and at zcap = 5 µm the averagefield is β = −0.021. Compared to zcap = 0 µm the magnitude of the average field fromthe beads is considerably increased by a 5 µm thick capping layer.

Concluding remarks concerning SU-8 design

A layer of SU-8 is needed for the binding chemistry, and an SU-8 layer yielding z0 = 0.5 µmis chosen as the sensor coverage. Summarizing the average field produced by a monolayerof beads covering the SU-8 layer gives: β1 = −0.006 (Design 1), β2 = −0.023 (Design 2),and β3 = −0.021 (Design 3). Based on the first two numbers, Design 2 of the SU-8 layeris chosen. Unfortunately, the idea of Design 3 has been introduced too late in the processto realize experimentally. The smaller β3 than β2 is fully compensated for by avoiding therisk of positive contributions.

4.6.1 Numerical values for Design 2

This subsection describes the SU-8 design chosen for bead detection experiments. Theaverage field produced by beads in the different parts of the sensor is evaluated separatelyfor each individual part. Contributions outside the sensor (Eo+F) are now included.z0 = 0.5 µm at the sensor (C+Ei), and z0 = R outside (Eo+F).

Table 4.6 reports the numerical values obtained for the SU-8 design chosen for assayexperiments. The last two lines separates the contributions outside the sensor area (Eo+F)from those inside the sensor area (C+Ei). The magnitude of these two contributions areclose to each other but the positive contribution from the sensor area dominates such thatthe total field becomes positive.

∆x′ ∆y′ A/Atot z0 (µm) Trp′ β monolayerC 0.2 2 0.24 0.5 -364.566 -0.0128Ei 0.1 1 0.06 0.5 -287.689 -0.0097Eo 0.1 1 0.051 0.125 +511.380 +0.0172F1 0.1 1 0.234 0.125 -66.085 -0.0022F2 0.1 1 0.166 0.125 +19.770 +0.0007F3 0.1 1 0.249 0.125 +289.328 +0.0102

C+Ei 0.2 2 0.3 0.5 - -0.0225Eo+F 0.2 2 0.7 0.125 - +0.0257

Table 4.6: Trapezoidal sums, Trp′, of the integrals, Int′ (non-normalized), and the averagemagnetic field, β = µ0〈Hx〉/B Eq. (4.31), produced by a monolayer of beads covering therespective area. The design of the SU-8 layer is shown in Fig. 4.10 (2).

4.7. CONCLUSION 51

Both extremes, beads covering only the sensor area or beads covering only the areaoutside the sensor, leads to large average fields of |β| ≈ 0.02. But a combination of thetwo would lead to a decrease in the average field. This average field will be approximatelylinear with the difference in coverage,

β = 0.0257 · ξoutside − 0.0225 · ξsensor ⇒ (4.33)

β ≈ 0.024(ξoutside − ξsensor)

where ξ is the bead coverage with respect to a closely packed monolayer of beads. Eq. (4.33)assumes non-interacting beads.

4.7 Conclusion

In summary, this theoretical analysis has shown that the planar Hall sensor is capable ofdetecting a single bead. For specific commercial bead products the sensor can be designedfor single bead detection based on the single bead S/N study presented.

The signal from a monolayer of closely packed beads is estimated for non-interactingbeads. Given the constraint that z0 = 0.5 µm at the sensor area, an SU-8 design, whereonly the sensor area is covered, should be optimal in terms of the average bead field,namely β = −0.023 (normalized to the applied field). If beads are spilled outside thesensor area, Eq. (4.33) approximates the normalized field average.

Chapter 5

Nickel planar Hall sensors

Having decided on the planar Hall sensor for magnetic detection in immunoassays, the firstexperimental test of the principle should be considered. In MIC’s clean room, the onlyavailable anisotropic magnetoresistive material is nickel, and obviously our first attempton constructing planar Hall sensors is with nickel as sensing material [2].

This chapter describes the fabrication and characterization of nickel planar Hall sen-sors. The aim is to verify that the sensing principle can be used for bead detectionexperiments.

5.1 Design

In this section the design of the nickel planar Hall sensor is described. The design in-cludes the appearance of the sensor, viewed in a microscope, the sensing material, and abiologically active material for attachment of biomolecules.

The design of sensor appearance has changed considerably during the work with thenickel sensors as can be observed by comparison of Fig. 5.1 and Fig. 5.2. The magneticstack for sensing is: Si-substrate/Ni(250A)/Au(20A), where the nickel layer is the activesensing material and the gold layer is a protective layer. Gold and SU-8 surfaces havebeen considered for attachment of biomolecules.

5.1.1 Sensor appearance

If a current runs through an AMR material, the planar Hall effect induces an electric fieldperpendicularly to the current, as calculated in Eq. (3.7). This sensing geometry suggestsa cross geometry of the sensor design as illustrated in Fig. 5.1.

The planar Hall sensor uses all its active area for detection, hereby is understood thatthe sensor will measure an average of the field felt by the total area in between the fourleads. This yields the possibility of tuning the design to match exactly the requirementsof an application. The first demonstration is detection of superparamagnetic DynabeadsM-280 with a diameter of 2.8 µm. Choosing a sensitive area of 20µm×20µm there will beroom for approximately 50 beads on top of the sensor, which gives a statistical probabilityof getting beads on the sensor by random methods.

53

54 CHAPTER 5. NICKEL PLANAR HALL SENSORS

Figure 5.1: Left: Micrograph of the first nickel planar Hall sensor. Yellow is metal, greenis the silicon wafer. The metal arrow shows the direction of the easy axis. The measuringgeometry is indicated in the figure, I is the current, and V the voltage. Right: Crosssectional view of the layer stack: substrate/Ni(250A)/Au(20A). Layer thicknesses are inA.

Figure 5.2: Left: Micrograph of a later design of the nickel planar Hall sensor. The goldleads are much closer to the sensor cross. Right: Cross sectional view of the magneticstack: substrate/Ni(250A)/Au(20A). Layer thicknesses are in A.

5.1. DESIGN 55

A picture of the first operational device is presented in Fig. 5.1. Only one sensor crosswith four bonding pads, two for current and two for voltage, is contained on the chip. Thesensitive area is 20µm×20µm.

During the nickel sensor project this first appearance has evolved into the sensor designshown in Fig. 5.2. The chip now contains three sensors, A, B, and C, which can be usedas references against each other. Furthermore, the electrical leads are moved closer to thesensor cross. This configuration ensures that the majority of the voltage drop lies acrossthe sensor cross. The active sensing area of 20µm×20µm is maintained.

5.1.2 Sensor material

An AMR sensor involves the material property, sensor geometry, and a detectable fieldrange. If the AMR effect is absent as in isotropic magnetoresistive materials it cannotbe used as a magnetic field sensor in this context. Most materials are actually isotropicwith respect to magnetoresistance, only ferromagnetic materials exhibit anisotropic mag-netoresistive effects. Ferromagnetic materials are, e.g., iron, nickel, and cobalt, or alloysinvolving these. Nickel has an AMR of approximately 2 % [6], which is deficient withrespect to permalloy that has an AMR of 5 %. These are bulk values, for thin films lessmust be expected. For permalloy films with a thickness of 20 nm, an AMR of 2.2 % hasbeen reported [33].

The sensitivity of the resulting sensor is theoretically inversely proportional to thethickness of the AMR layer, indicating that the thinnest layer possible should be chosenfor the devices. It is, however, unfeasible to make the layer infinitely thin, because at somepoint the AMR effect disappears. An investigation of AMR as a function of permalloythickness can be found in Ref. [33]. A relatively steep rise is reported for thicknesses up to20 nm after which the curve flattens considerably. If nickel behaves similarly, thicknessesof 20-25 nm are optimal.

Nickel has a low electron affinity and corrodes easily. For protection against corrosion a2 nm gold layer is deposited on top of the nickel layer as proposed in Ref. [1]. Additionally,since it is conducting, protection with gold ensures electrical contact to the bonding pads.

Montaigne et al. [1] constructed an easy axis of magnetization by step bunching asilicon wafer misoriented with respect to the [111] crystal direction. The reported magneticmoments of the film are thus oriented parallel or perpendicular to the steps depending onwhich material is deposited. In the case of permalloy, the easy axis is induced parallel tothe steps [1].

In stead of step bunching we use an applied magnetic field during deposition. Themagnetic field is produced by a magnetic frame, see drawing in Fig. 5.3. The frame isconstructed of two bars of soft iron held together by permanent magnets. The mag-netization of the two arrays of permanent magnets are aligned parallel resulting in twoadjacent loops of magnetic flux as sketched in the figure. The sketch is oversimplified butgives an idea of how the principle of flux-closure creates a relatively homogeneous field inthe gap of the magnetic frame. This configuration produces a magnetic flux density ofB = 8 mT in the center of the frame. When experiencing a field during deposition, theminimum energy configuration of the ferromagnetic material depends on the field. The


Figure 5.3: Schematic drawing of the magnetic frame used during Ni deposition. It pro-duces a magnetic flux density of B = 8 mT in the center of the frame. The field lines areshown as arrows. Left: Frame on top of wafer. Right: Frame viewed from above.

crystal structure adapts in accordance with this energy minimization and thus experiencesuniaxial anisotropy.

5.1.3 Biologically active material

For attachment of biomolecules, three approaches have been considered, though no actualbio-experiments have been performed with the Ni sensors.

1. Attachment to gold

2. Attachment to gold on SU-8

3. Attachment to SU-8

The first attempt of attaching thiolated DNA to the thin gold layer covering the sensoris unsuccessful because the gold layer is too thin to be properly activated. The secondattempt is to separate the sensor and a much thicker gold layer by 0.5 µm SU-8. It isdifficult to adhere the gold to the SU-8 surface. At this point, another approach is pursued,namely attaching DNA directly to SU-8, which in the successful case will facilitate notonly the functionalization of planar Hall sensors but also other micro-fabricated sensors[34]. During this study, SU-8 reveals a great binding capacity towards both DNA andproteins. The details on attachment of biomolecules to the SU-8 surface will be given inChapter 7.

5.2 Nickel sensor fabrication

This section describes the fabrication procedure for nickel planar Hall sensors. It is adescription of the initial clean room work leading to a conference proceeding [2] andsupporting the design and choices for the exchange biased sensors. The nickel planar Hallsensors are fabricated on standard silicon wafers without electrical insulation.

5.2. NICKEL SENSOR FABRICATION 57

5.2.1 Deposition of magnetic Ni films

For deposition of nickel films an e-beam deposition system, Alcatel Model SCM600, is used.In the e-beam system, the metal is placed in a crucible and heated by an electron beamextracted from a tungsten filament. Due to the beam profile, heating can be controlledlocally in the crucible and sublimation is achieved without melting all the metal. Thisreduces cross contamination from the crucible or other metals placed in different crucibles.Four different materials can be deposited without breaking the vacuum, hence we are ableto deposit first the Ni sensing layer and then the Au protective layer.

The deposition is controlled by manually opening and closing a shutter between themetal crucible and the wafer. Since our metal layers are very thin compared to normalstandards, we use lower deposition rates in order to facilitate the opening and closing ofthe shutter. The layer thickness of deposited material is measured during deposition by aquartz crystal that changes its resonance frequency proportionally to the mass depositedon top of it.

A consequence of the deposition method is a high probability of obtaining polycrys-talline nickel films. Hence, the nickel films will exhibit magnetic domains. This is incontrast to epitaxial films, which are more likely to exhibit single domain behavior.

5.2.2 Photolithography

Sensors are built on top of a silicon wafer using conventional photolithography. Thephotoresist, which is a polymer that changes its chemical properties when exposed toultraviolet light, is used to create microscopic structures of resist on top of the wafer. Apositive resist becomes soluble in a solvent, called the developer, when exposed to UVlight, and a negative resist becomes insoluble in the developer.

First the photoresist is spun on to the wafer, the spinning creates a uniform layer ofthe resist. Then the resist is exposed to UV light through a mask. The mask is a glassplate transparent to UV light with chromium deposited at predefined patterns in order toavoid resist exposure here. The patterns, e.g. corresponding to sensor crosses, are thustransferred to the photoresist. When the wafer is subsequently flushed with developer, thesoluble part of the resist is dissolved in the liquid, and the transferred pattern remains.

This mould of photoresist can be used for general purposes. One can etch away materialin the bottom of the grooves defined by the resist, or one can fill the grooves with materialand build up structures. If we choose to deposit a layer of metal on top of all and dissolvethe photoresist in acetone, this type of process is called a lift-off. The remaining patternof metal will be identical to the original pattern of the mask.

5.2.3 Process steps

Patterning the nickel-gold film for sensors and the thicker gold film for leads is done bylift-off. The negative imaging exposure process is used to facilitate the lift-off. Followingclean room processing, the devices are cut from the wafer with a dicing saw and wirebonded to a chip carrier constructed of an Al2O3 substrate and 14 CrAu bonding pads.


Figure 5.4: Lithography steps for nickel sensor. 1. The wafer is covered in photoresist.2. Structured resist. 3. Deposition of Ni/Au film. 4. After lift-off the sensor patternremains.

• A 4” silicon wafer is used as substrate. Typical resistivity ∼ 1 − 20 Ω cm. Surfaceoxide is removed in HF buffer for better adhesion of photoresist.

• 1.5 µm photoresist AZ5214e is spun onto the wafer. Use SSI System 150 Track 1.Conditions: program PR1-5.

• Sensor crosses and leads are defined in the photoresist by exposure through a mask.Use negative imaging procedure. Equipment: SUSS Mask Aligner MA6/BA6. Con-ditions: exposure dose 9 mW cm−2. Negative procedure: 8 s exposure with mask, 2min bake at 120 oC hotplate, 25 s flood exposure (without mask).

• Development in AZ531. Conditions: time: 55 s, stirring of solution.

• Nickel sensing layer and protective gold layer is deposited by e-beam evaporation inAlcatel Model SCM600. Stack: substrate/Ni(25nm)/Au(2nm). Conditions: back-ground pressure: 10−6 mbar, processing pressure: ≈ 3 · 10−6 mbar, RF voltage: 8kV, deposition rates: 0.1 nm/s.

• Lift-off in acetone. Conditions: 30-60 s ultrasonic agitation.

• 1.5 µm photoresist is spun on, patterned into bonding pads by negative imagingprocess, and developed. Conditions as above.

• Deposition of 0.5 µm Au for current/voltage leads and bonding pads in AlcatelModel SCM600. Conditions: background pressure: 10−6 mbar, processing pressure:≈ 5 · 10−5 mbar, RF voltage: 8 kV, deposition rate: 1 nm/s.

• Lift-off in acetone. Conditions: ultrasonic agitation. Ultrasonic agitation can beassisted by heating the acetone bath prior to stripping of the resist.

• Dicing with a Tempress Dicing Saw, and wirebonding using an ultrasonic wirebonder.

The lithography steps for fabricating a nickel planar Hall sensor are shown in Fig. 5.4.

5.3. EXPERIMENTAL 59

Figure 5.5: Schematical drawing of the electronic setup for Ni sensor characterization. 1:Helmholtz coils, 2: power supply, 3: sensor box, 4: chip with three sensors A, B, C, 5:lock-in amplifier, 6: computer.

5.3 Experimental

In this section the experimental setup constructed for Ni sensor characterization is de-scribed. It consists of magnetic field control, a sensor box, and sensor current controlcombined with voltage measurement.

5.3.1 Experimental setup and conditions

The experimental setup for nickel planar Hall sensors is presented in Fig. 5.5. Electricalconnections are shown as solid lines, and data transfer is indicated with an arrow. Belowis a description of the components shown in the drawing.

Magnetic field control

Two methods of producing magnetic fields have been conducted in the experiments with Nisensors. One is a permanent magnetic frame much like the one used for nickel deposition(Fig. 5.3), only this frame is larger and produces a flux density of B = 42 mT. Using thisframe, the sensor is always saturated along the direction of the external field.

The other method is to produce the field with a set of Helmholtz coils. The Helmholtzcoils consist of two identical coils placed with their centers at a distance identical to thecoil radius. This configuration ensures that the second derivative of the field vanishes inthe point just between the two coils. Therefore a set of Helmholtz coils produces a veryhomogeneous field in the central volume between the two coils. A picture of the Helmholtzcoils can be found in Fig. 6.6, number 1.


1. Helmholtz coils. The variable applied field is supplied by a set of homebuilt Helmholtzcoils [3]. Due to heat generation the Helmholtz coils are suitable for use with cur-rents of no more than IDC = 2 A and, when using this value, only for a few minutes.The calibration of the Helmholtz coils is 7.57 mT A−1 - 8.48·10−3 mT [3]. Thiscalibration, carried out in Ref. [3], is later found to be 2.8 % too low. Chapter 6presents a more accurate calibration.

2. Midec Type SK 150-2 power supply. Supplies the current for the Helmholtz coils.The power supply is controlled via LabView [35].

Sensor box

3. Sensor box. The sensor box is mounted on a rotatable table with a protractor. Thisensures that the angle between the applied field and the current can be controlledwith ±0.5 o resolution.

4. Chip with three sensors, A, B, and C. The user can switch between the three sensorsmanually by interchanging the cables.

Sensor current control and voltage measurement

5. Stanford Research Systems Model 830 lock-in amplifier [36]. The current through thesensor is controlled by the sine output from the lock-in amplifier. The modulationfrequency is f = 1.104 kHz. A VAC = 1.004 V (maximum amplitude) sine functionpassed through a Re = 4.8 kΩ shunt resistor in series with the sensor, Rsensor = 200Ω (measured with a multimeter), gives a current of IAC ≈ 200 µA.

The lock-in amplifier measures the voltage output from the sensor using the sineoutput as reference. Lock-in settings are given in Table 5.1.

Experiment control

6. Computer. The experiments are controlled via LabView [35].

Vout (V) f (kHz) Display Sensitivity (µV) Signal1.004 1.104 X, Y 50 A-Bτ (s) Filter slope (dB) Reserve Line filters Sync filter

1 12 low noise off on

Table 5.1: Lock-in amplifier settings for Ni sensors. Vout is the output voltage from thelock-in amplifier (maximum amplitude) with frequency f . τ is the time constant.

5.4. RESULTS AND DISCUSSION 61

5.4 Results and discussion

This section presents the results obtained for the nickel sensors. First the angular de-pendence of V (φ) in Eq. (3.8) is investigated. Then the sensor response versus appliedfield is presented and discussed. Finally bead detection experiments are quoted from theconference proceeding [2].

5.4.1 Sensor response vs. angle

In order to investigate the sin(2φ) relation of Eq. (3.8) the sensor is saturated by the strongmagnetic frame and the voltage drop is measured as a function of angle between the appliedcurrent and the saturation magnetization. No shunt resistor is included in the electricalcircuit. Measurements of this dependence is shown in Fig. 5.6 for a sensor constructed of25 nm thick nickel. A sin(2φ) behavior is observed as theoretically predicted.

5.4.2 Sensor response vs. field

Procedure

Prior to each measurement, the magnetization of the sensor is saturated along the easyaxis. For this procedure, the Helmholtz coils are used. The table is turned such thatthe field is aligned along the sensor’s easy axis. Then the current through the coils isset at its maximum value and subsequently slowly reduced. This step ensures that themagnetization always has the same starting point. After resetting the magnetization alongthe easy axis, the table is turned 90 o. Now the applied field is aligned perpendicularly tothe easy axis.

Results

Measurements of the sensor response as a function of applied field is presented in Fig. 5.7for a nickel sensor of 20 nm thickness. This graph is typical for the nickel sensors. Observedis first a linear region as a function of increasing applied field, then a maximum is reachedand afterwards a drop in signal in accordance with theory. When decreasing the field afterreaching B = 1.75 mT, almost no signal response as a function of field is observed.

First the linear part of the curve is considered. From fabrication in the clean room, themagnetic moments of the nickel atoms are aligned in the direction of the applied field fromthe small magnetic frame. When an external field subsequently is applied perpendicularlyto this direction, the moments rotate away from their original alignment and the sensorvoltage changes in accordance with theory. When the moments are at 45 degrees to thecurrent, the sensor signal attains its maximum value at B = 1.3 mT. For larger fields, andhence larger angles, the signal drops again.

At sensor saturation (not reached in Fig. 5.7), the magnetization will be fully alignedwith the applied field, and when the applied field is subsequently reduced, the momentsrotate back into the direction of the easy axis. However, at this point, no distinctionbetween the two directions of the easy axis is possible. Due to this fact, the magnetic


0 20 40 60 80 100 120 140 160 180

-20

-15

-10

-5

0

5

10

15

20

Vac

= 0.050 Vf = 1.104 kHzB = 42 mT

Sensor: 25 nm Ni

S

igna

l (V

)

Angle to field (degs.)

Figure 5.6: Angle dependence of the Ni planar Hall sensor. Experimental conditions areon the graph, no shunt resistor is in the electric circuit. The sensor is saturated (B = 42mT) and the signal is measured as a function of angle between current and a saturatingapplied field.

0.0 0.5 1.0 1.5 2.010

12

14

16

18

20

22Sensor: 20 nm Ni

Sig

nal (

V)

Applied field (mT)

Figure 5.7: Signal vs. applied field for a nickel planar Hall sensor of 20 nm thickness.Experimental conditions are: a modulating frequency of f = 1.104 kHz, and a current ofIAC = 200 µA.

5.4. RESULTS AND DISCUSSION 63

moments will distribute themselves randomly into the two antiparallel directions. Thesedirections are magnetically equivalent, but the resulting electrical signals are of oppositesigns. On average, the signal will be zero, or significantly reduced as observed in thegraph. When sweeping the field forth and back between higher and higher values ofmagnetic fields, the effect of this redistribution of magnetic moments is enhanced as afunction of max field.

What is also clear in Fig. 5.7 is that the zero applied field signal at the start of themeasurement does not coincide with the signal at the end of the measurement. Thereason for this observation is probably a temperature drift. A temperature drift towards asaturation value is observed in many measurements, and the sensors are usually left withapplied current for an hour prior to the experiment in order to obtain an equilibrium beforestarting the measurements. Another explanation could be remanence in the magnetizationstate of the sensor.

Due to the magnetic equivalence of the two antiparallel states, the sensor has to bereset prior to each measurement. By reset, is meant, that all the magnetic momentsare aligned along one direction by applying a magnetic field parallel to the easy axis asdescribed above. When this is taken care of properly, the sensors show stable responses.For the planar Hall sensors of 20 nm thick nickel typical sensitivity values are S0 ≈ 35µV mT−1 mA−1, which are found as the slope of the signal versus applied magnetic fieldcurve for small applied fields for repeated measurements of several sensors.

5.4.3 Bead detection experiments

Procedure

The magnetization is set along the easy axis as described in the procedure for sensorresponse vs. field.

Additionally, the sensors are left for minimum one hour with applied current priorto each measurement. The drift thus stabilizes at an equilibrium value, which can besubtracted from the measurements.

1. The slope of the sensor response as a function of applied magnetic flux density ismeasured under controlled experimental conditions. The current through the sensoris IAC = 200 µA, the modulation frequency is f = 1.104 kHz, the signal is measuredusing lock-in technique, and a large off-set is subtracted from the measurements.

2. A droplet containing beads is placed on top of the sensor using a syringe. Thedroplet is left to dry, and beads sediment on top of the sensor.

3. The slope is measured under the same experimental conditions as the first slopedetermination. Comparison of the two slopes give the signal from the beads.

Results

Fig. 5.8 shows bead detection measurements with the nickel planar Hall sensor presentedat the Eurosensors 2003 conference [2]. Comparison of the two responses before and after


0.0 0.1 0.2 0.3 0.4 0.5 0.60

1

2

3

4

5

Sensor: 20 nm Ni

slope=3.5 V/mT

slope=7.0 V/mT

Signa

l (V)

Applied field (mT)

no beads with beads (meas. 1) with beads (meas. 2)

Figure 5.8: Bead detection with the nickel planar Hall sensor. Left: Electrical measure-ments of signal versus applied field. Experimental conditions are: a modulating frequencyof f = 1.104 kHz, and a current of IAC = 200 µA. The slope is reduced by the presence ofbeads. Right: Micrograph of the sensor with Dynabeads M-280.

beads have been placed on the sensor reveals the presence of the beads. The sensor signalversus applied field slope is reduced by the presence of beads. Visual inspection in themicroscope shows adhesion of at least six Dynabeads M-280 within the sensitive area ofthe sensor, and more are present in close vicinity. After washing off the beads with water,the field response is restored to match the first experiment.

5.5 Conclusion

This chapter has shown that the planar Hall effect can function as detection principle formagnetic bead sensors. The angular dependence of a saturated sensor obeys the sin(2φ)relation of Eq. (3.8). The sensitivity is S0 = 35 µV mT−1 mA−1, when the magnetizationstarting point of the sensors is specified prior to each measurement. Furthermore, detectionof Dynabeads M-280 has been realized using the nickel planar Hall sensors.

An obvious deficient property of the nickel planar Hall sensor is the need for restoringthe magnetization along one of the two antiparallel directions of the easy axis prior to eachmeasurement. Another issue is the magnitude of anisotropic magnetoresistance, which islarger in permalloy. One of the characteristic properties of permalloy is, however, the verysmall anisotropy energy, which is the cause of the sensor saturation in the Earth’s magneticfield experienced by Montaigne et. al. [1]. Nickel does not suffer from this deficiency.

One unsolved item is the temperature dependence of the sensor. When measuring thesignal as a function of time, the signal is observed to drift from its initial value and approach

5.5. CONCLUSION 65

an equilibrium value. This drift of the signal could be related to the actual temperature ofthe sensor during the measurements. When the sensor is heated above room temperature,the signal rises, and when the sensor subsequently is cooled below room temperature, thesignal drops below the ”steady-state” value corresponding to room temperature. Thisobservation will be addressed later along with the results of the permalloy sensors, whichalso exhibit temperature drift.

Finally, during the study of nickel planar Hall sensors, it has been found that SU-8 isan excellent material for attachment of biomolecules [34], see Chapter 7, where the resultsfrom this study are described.

Chapter 6

Exchange biased permalloy planarHall sensors

Figure 6.1: Final chip, Batch 2. The chip consists of three sensors, A, B, C, where B isthe electrical reference, and A and C are used for bead detection experiments.

This chapter describes the fabrication and characterization of exchange biased permal-loy planar Hall sensors. The reason for choosing permalloy as sensor material is the higherAMR compared to Ni [6]. Additionally, a well-defined starting point of the magnetizationcan be induced by exchange coupling the permalloy layer to an antiferromagnet [27].

The exchange coupling introduces a unidirectional anisotropy in addition to the uni-axial anisotropy of the easy axis. This means that the magnetization prefers one of thetwo directions along the easy axis instead of equal preference. Hence, the magnetization

67

68 CHAPTER 6. EXCHANGE BIASED PERMALLOY PLANAR HALL SENSORS

Figure 6.2: Sensor design, Batch 1. The central cross (brown), which is the active sensingarea, is 10µm×10µm. The leads (white) are made of Al.

state is always well-defined in zero applied field, and there will be no need for magnetizingthe sensor along the easy axis as is the case of nickel sensors.

Fig. 6.1 shows a picture of the final chip with three exchange biased permalloy planarHall sensors used for immunoassay detection experiments. Planar Hall sensors have lowersignal magnitude than existing magnetic field sensors [26], however the low noise level ofthe planar Hall sensor makes it theoretically superior to other types of AMR sensors interms of the DC bead detection limit. Another beneficial characteristic of AMR sensorsis the easy fabrication method, which enhances the speed of fabrication, limits the cost ofa device, and promotes integration with micro-analysis biochips. The planar Hall sensoris no exception, though the exact thicknesses of the respective layers are less critical thanfor GMR and spin-valve sensors.

6.1 Design

This section presents the design of the exchange biased permalloy sensors. The designinvolves the sensor appearance and the sensor material as well as the design of the SU-8layer used for binding chemistry, which has already been discussed in Chapter 4.

Two permalloy planar Hall sensor batches have been fabricated, Batch 1 and Batch 2.

6.1.1 Sensor appearance

Batch 1

In Fig. 6.2 the resulting appearance of a Batch 1 planar Hall sensor is shown. The totalfabrication of Batch 1 took place at Institute of Engineering of Systems and Comput-ers - Microsystems and Nanotechnology (INESC-MN), Portugal, and a chip design sim-ilar to their spin-valve chips [9, 16] was used. Sensor crosses of active area 5µm×5µm,

6.1. DESIGN 69

10µm×10µm, and 20µm×20µm were fabricated, in the figure one 10µm×10µm sensor isshown. Each final chip consists of 10 sensors, each with four electrical leads all gatheredat a 1 cm2 chip area.

Batch 2

Fig. 6.1 shows the final chip design of Batch 2 with three planar Hall sensors on each chip.Three sensors are needed for an immunoassay experiment, one for the electrical reference,one for the sample, and one for the negative control. The sensor design of Batch 2 isidentical to that of the final design for Ni sensors (Fig. 5.2). The active sensing areas ofthe planar Hall sensors has been chosen to be 20µm×20µm.

6.1.2 Sensor material

The aim is a permalloy sensor experiencing unidirectional anisotropy. The easy axis isconstructed by applying a magnetic field during deposition. An easy direction is preferredinstead of an easy axis. This is accomplished by exchange coupling the magnetic layerto an antiferromagnetic layer, see Eq. (3.16) and Ref. [27]. The exchange bias ensures aunique starting point for the magnetization in zero applied field. Permalloy has a higherAMR than nickel. However, one of the characteristic properties of permalloy is the verysmall anisotropy energy, which is the cause of the sensor saturation in the Earth’s magneticfield experienced by Montaigne et. al. [1].

Fortunately, the two properties are coupled in a way that facilitates the modulation.In order to introduce a unique starting point for the magnetization, the ferromagneticfilm of sensing material is coupled to an antiferromagnetic film by exchange coupling. Theexchange energy is visualized by a shift of the easy axis loop to a lower field value in amagnetization versus applied field strength experiment. Since the energy of the permalloyfilm depends on both the anisotropy energy and the exchange energy, the rotation of mag-netization away from the easy direction depends on these two contributions as calculatedin Eq. (3.13) and Eq. (3.17).

Exchange coupling of the ferromagnetic film to an antiferromagnetic film thus improvesthe planar Hall sensor in three ways. First, the easy axis is substituted with an easydirection. Second, the exchange coupling introduces the possibility of using permalloy,which has superior magnetic properties compared to nickel. Third, the operation fieldrange has been extended also due to the exchange energy.

In Ref. [33] an investigation of AMR as a function of permalloy thickness is performed.A relatively steep rise is found for thicknesses up to 20 nm after which the curve flattensconsiderably. This indicates that thicknesses of 20-25 nm are optimal with respect toAMR.

Batch 1

As sensing material a 20 nm thick layer of Ni80Fe20 (permalloy) is chosen. For the antifer-romagnetic coupling a 20 nm thick Mn74Ir26 layer is used. There are several reasons forthe choice of layer thicknesses. In the case of Ni80Fe20, Fig. 4 in Ref. [33] shows that the


Figure 6.3: Cross sectional view of the IBD deposited stacks. Both deposited on anoxidized Si substrate. The sketch to the left shows the stack used for Batch 1 (bottom-pinned), and the sketch to the right is Batch 2 (top-pinned). Layer thicknesses are inA.

maximum AMR is reached at 20 nm thickness. In the Mn74Ir26 case, an exchange field ofHE = 30000 A m−1 is obtained for MnIr thicknesses down to 6 nm, when the permalloylayers are 3 nm thick [33]. It is, however, not advisable to use very thin layers due to lackof thermal stability, and based on the results of Ref. [33] 20 nm MnIr is thought to be agood compromise between thermal stability and AMR.

The magnetic dipole field, which is expected from a magnetic bead, see Eq. (A.17),falls off rapidly with distance, theoretically as the inverse cube of the distance. Thereforea sensing layer as close to the bead as possible is desired. 20 nm is, of course, negligiblein comparison with a 2 µm bead, but possible future applications involves measurementswith 50 nm beads, and sooner or later the spacing of the antiferromagnetic layer will playa role. Keeping this in mind, the ideal layer configuration is with the pinning layer belowthe sensing layer.

As adhesion layer, and for protection against corrosion, 3 nm thick Tantalum is used.On top of Ta, permalloy grows epitaxially, but in order to obtain MnIr single crystalgrowth a seed layer of CrNiFe or permalloy is needed. During my stay at INESC-MNCrNiFe was unavailable, and 5 nm permalloy was used as seed layer for MnIr. This resultsin a second loop in the magnetization versus applied field experiments presented below,and complicates theoretical investigation of the sensor field response.

The layer configuration of Batch 1 is: oxidized Si substrate/Ta(30A)/NiFe(50A)/MnIr(200A)/NiFe(200A)/Ta(30A).

Batch 2

Due to the problems of depositing a proper seed layer for MnIr, a top-pinned stack is chosenfor the second sensor batch. Batch 2 is thus: oxidized Si substrate/Ta(30A)/NiFe(200A)/MnIr(200A)/Ta(30A).

The cross-sectional view of the two stacks of Batch 1 and Batch 2 are shown in Fig. 6.3.

6.2. CLEAN ROOM FABRICATION OF EXCHANGE BIASED SENSORS 71

6.1.3 Biologically active material

Batch 1

The total chip of Batch 1 is protected by a 0.2 µm silicon oxide layer. This is also usedfor immobilizing biological material on top of the sensors [9, 16, 18]. The drawback ofsilicon oxide as a surface for biomolecular attachment is that the activation of the surfaceinvolves toxic chemicals. For large scale production of chips, toxic chemicals should beavoided due to environmental problems. For research applications the complications areless severe, but careful handling is mandatory.

Batch 2

In Batch 2, SU-8 is used to bind antibodies on top of the sensor, see Chapter 7 for details onthe immobilization procedure. The SU-8 can be made as thin as necessary, and a fortunatething about SU-8 as immobilization surface is that all chemical attachment steps can bemade using a PBS solution (phosphorous based saline), which is both cheap and non-toxic.

At the time of processing, it was not known that antibodies stick as well to the residualSU-8 layer left after development. Therefore, 0.5 µm SU-8 pads are made on top of thesensors as illustrated in Fig. 6.4. The SU-8 pads are introduced to immobilize antibodiesonly on top of the sensor. The aim is to avoid loss of antigen at non-active areas of thechip. However, this is an unfortunate design with respect to bead signal from a monolayer,where both the sensor and the area outside the sensor is covered with beads. As discussedin Chapter 4, the monolayer signal is proportional to the difference between the coverageof beads inside and outside the sensor, Eq. (4.33). If the coverage is approximately equalinside and outside the sensor area, the resulting signal will approach zero.

6.2 Clean room fabrication of exchange biased sensors

In this section the fabrication process of the exchange biased sensors is described fromwafer to the final chip seen in Fig. 6.1. The photograph shows the Batch 2 chip mountedon a chip carrier.

The processing involves deposition of the magnetic films and subsequent structuringof the films into sensors. Then electrical leads are constructed, and the biologically activematerial is deposited. Packaging involves mounting the diced chip on a carrier and wirebonding.

6.2.1 Deposition of magnetic thin films

The magnetic films for planar Hall sensor fabrication are deposited in collaboration withINESC-MN. The equipment used for deposition is the Nordiko N3000 ion beam systemdescribed in Ref. [33]. In comparison with other thin film techniques, where sputteringis the closest competitor, the ion beam deposition (IBD) technique offers good control ofdeposition conditions, such as low deposition rates and low processing pressures. The base


pressure in the IBD chamber is below 7 · 10−6 Pa, and during operation the pressure risesto 5.5 · 10−3 Pa.

In Ref. [33], a description of the system operation is provided along with an examina-tion of the level of cross-contamination from other targets and the shield. Contaminationof the deposited films is approximately 400 ppm of which metallic impurity is below 140ppm. The operation conditions are optimized in terms of the beam profile, and the re-sulting uniformity of the film thickness is ±1 % across a four inch wafer. The substrate iswater cooled during deposition, which is especially important for ion milling processing,where considerable excess heat is generated. A permanent magnetic frame provides 3200A m−1 magnetic field strength during deposition. The easy axis can be aligned with ±2o accuracy.

Deposition is achieved by bombarding ions into the target. Xenon ions are extractedfrom the deposition gun by an acceleration voltage of -300 V and hit the target withan energy of 1450 eV. Target material is sputtered off the target and thus deposited onthe substrate wafer. The beam current is set by the RF power to 29 mA. The resultingdeposition rates are Ta 0.20 A s−1, Ni80Fe20 0.32 A s−1, and Mn74Ir26 0.31 A s−1.

6.2.2 Process steps Batch 1

Fabricating the planar Hall sensors of Batch 1 [4, 5] is a process sequence with threemask steps, four deposition steps, two etch steps and a lift-off, plus the intermediatepreparations.

• A 3” silicon wafer is insulated electrically by a sputtered 0.5 µm Al2O3 layer. Ontop 15 nm antireflecting TiWN2 is sputtered.

• The sensor material is deposited using ion beam deposition (IBD) Nordiko 3000. Thestack is: substrate/Ta(30A)/NiFe(50A)/MnIr(200A)/NiFe(200A)/Ta(30A). Deposi-tion conditions: Substrate rotation 50 %, substrate pan: 80 o, magnetic field duringdeposition: 40 Oe, beam current: 29 mA.

• 1 µm photoresist is spun on. Pre-treatment: Vapor prime (5’). Conditions: Speed3500 rpm.

• Sensor crosses are patterned using direct laser writing. Conditions: exposure power:82 mW.

• Mask development. Conditions: Developer 5/2.

• The excess of material is etched away by ion milling leaving sensor crosses. Etch inNordiko 3000. Conditions: RF power: 64 W, RF cycle: +488.5V/-194V, +29.5mA/-1.6mA, stabilization plasma: 8 sccm Ar, substrate rotation: 40 %, substrate pan:70 o, chamber pressure: 5.4 · 10−7 Torr, total etch time: 600 s.

• Resist stripping in Microstrip 2001. Conditions: 80 oC, ultrasonic agitation.



• Electrical connections are patterned using direct laser writing. Conditions: exposurepower: 82 mW.


• 0.3 µm Al for electrical leads and 15 nm TiWN2 protective layer is deposited inNordiko 7000 sputter system. Conditions: Clean machine before run. Surface etchconditions: sputter etch 1: 40 s, 70 W, sputter etch 2: 40 s 50 W, stabilizing plasma:50 sccm Ar. Al deposition conditions: total time: 80 s, DC power: 2 kW, stabilizingplasma: 50 sccm Ar. Protection layer conditions: total time: 27 s, DC power: 0.5kW, stabilizing plasma: 50 sccm Ar + 10 sccm N2.

• Electrical connections are defined by lift-off of the photoresist and the metal adjacentto the leads. Conditions: Microstrip 2001, 80 oC, ultrasonic agitation.

• 0.2 µm silicon oxide is sputtered on top the wafer. The oxide is used for surfaceprotection and for immobilizing biochemical substances.


• Mask patterning of holes for wire contacts using direct laser writing. Conditions:exposure power: 82 mW.


• Reactive ion etching through the oxide to open contacts for wire bonding.

• Dicing and wire bonding remains to give the final functional chips.

6.2.3 Process steps Batch 2

Fabricating the planar Hall sensors of Batch 2 is a simple process sequence with onlytwo mask steps, three deposition steps, one etch and a lift-off step, plus the intermediatepreparations. However an additional mask step is introduced in order to place SU-8 ontop of the sensors. The SU-8 functions as adhesion layer for biochemical species.

• A 4” silicon wafer is insulated electrically by a thermally grown 0.5 µm SiO2 layer.

• The sensor material is deposited using ion beam deposition (IBD) with depositionconditions as in Batch 1. The stack is: substrate/Ta(30A)/NiFe(200A)/MnIr(200A)/Ta(30A). Deposition conditions: Substrate rotation 50 %, substrate pan: 80 o,magnetic field during deposition: 40 Oe, beam current: 29 mA.

• 1.5 µm photoresist AZ5214e is spun onto the wafer. Equipment: SSI System 150Track 1. Conditions: program PR1-5.


• The sensor crosses are defined in the photoresist by exposure through a mask. Equip-ment: SUSS Mask Aligner MA6/BA6. Conditions: time 10 s, exposure dose 9 mWcm−2.

• Development in AZ531. Conditions: time 55 s, manual stirring of solution.

• The excess of material is etched away by ion milling leaving sensor crosses when thephotoresist is removed. Machine: Commonwealth IBE system located at UniversitatHanover, Institut fur Mikrotechnolgie, Germany.

• 1.5 µm photoresist AZ5214e is spun on, and the resist is structured with grooves inthe pattern of the electrical connections. Same spin, exposure dose, and developmentconditions as above. Exposure conditions: Negative imaging procedure: 8 s exposurewith mask, 2 min bake at 120 oC hotplate, 25 s flood exposure (without mask).

• 0.3 µm Au is deposited by e-beam evaporation in Alcatel Model SCM600 e-beamsystem. Conditions: background pressure: 10−6 mbar, processing pressure: ≈ 5·10−5

mbar, RF voltage: 8 kV, deposition rate: 1 nm/s.

• Lift-off in acetone. Conditions: ultrasonic agitation.

Immunoassay SU-8

• The sensors produced for immunoassays are subjected to another mask step. Firstthe wafer is dehydrated. Equipment: HMDS oven, IMTEC star2000. Conditions:time 1 hour, temperature 100 oC, vacuum (no HMDS vapor).

• The wafer is covered with very thin SU-8 photoresist (0.5 µm). Equipment: KarlSuss RC8-THP spin coater. Conditions: 2000 rpm.

• Prebake. Equipment: Karl Suss RC8-THP spin coater. Conditions: hotplate set to90 oC. Bake sequence: 30 s at 5 mm distance above the spinner hotplate, 30 s at2 mm distance, 2 min directly on hotplate, 30 s at 2 mm distance, 30 s at 5 mmdistance.

• Photolithography to structure the SU-8 into small squares covering only the sensingarea of a sensor, that is the central 20µm×20µm. Equipment: SUSS Mask AlignerMA6/BA6. Conditions: time 30 s, exposure dose 9 mW cm−2.

• Postbake. Equipment: Karl Suss RC8-THP spin coater. Conditions: hotplate set to90 oC. Bake sequence: as prebake.

• Development: 1 min in clean PGMEA (1-methoxy-2-propyl acetate). Rinse brieflyin isopropanol. Dry with N2 gas.

• Dicing and wire bonding remains to give the final functional chips.

A schematic drawing of the process sequence is given in Fig. 6.4 along with photos ofthe sensors at the various stages.


Figure 6.4: Lithography steps Batch 2. Left: schematic drawing of the lithography process,cross sectional view. Right: The top drawing is the cross section of the deposited material,the others are pictures of the wafer when processing moves along, the pictures are viewedfrom above. The picture sequence is: Structured photoresist, Dry-etch patterned sensors,Au leads, and SU-8 pads.


6.3 Experimental

In this section the experimental characterization equipment is described. First the VSMused for magnetic characterization of the films. Then the electrical setup used for electricalcharacterization of the Batch 2 sensors and bead detection measurements with Batch 2sensors. The electrical setup consists of magnetic field control, temperature control, sensorcontrol, and sensor output measurement.

Finally, the experimental conditions for Batch 2 sensors are described. The lock-inamplifier settings for noise characterization, sensor sensitivity, and biological detectionconditions are summarized in Table 6.1.

6.3.1 Vibrating sample magnetometry

The magnetization curves of the IBD deposited films are investigated using a vibratingsample magnetometer (VSM), Lakeshore Model 7407. The sample is placed in the center ofa detector coil assembly and two electromagnetic coils producing a variable magnetic field.The sample is set to vibrate at a specific frequency which induces an electromotive force inthe detector coil assembly due to Faraday induction. The voltage is directly proportionalto the magnetization of the sample. Hence the absolute magnetic moment can be obtainedafter calibration with a known sample. Sweeping the magnetic field and measuring themagnetization of the sample as a function of applied field reveals the magnetic propertiesof the sample.

6.3.2 Electrical characterization setup, Batch 2

Fig. 6.5 shows a schematic drawing of the experimental setup constructed for sensor char-acterization. Electrical connections are shown as solid lines, dashed when an optional pathcan be chosen, and data transfer is indicated with an arrow. Below is a description of thecomponents shown in the drawing. Fig. 6.6 is a photograph of the experimental setup,and the picture shown in Fig. 6.7 (left) reveals the inside of the sensor box. Fig. 6.7 (right)is a schematic drawing of the setup controlling the temperature of the chip.

Magnetic field control

1. Helmholtz coils. The variable applied field is supplied by a set of homebuilt Helmholtzcoils [3]. Due to heat generation the Helmholtz coils are suitable for no more thanI = 2 A and, when using this value, only for short periods of time. The calibrationof the Helmholtz coils is 7.7819(0.020) mT A−1 + 0.024(0.025) mT.

2. Keithley Model 6221. The current through the coils is controlled by a Keithley Model6221 high precision AC/DC current source. The source can deliver only I = 100 mADC, and hence the risk of over heating the coils is avoided when using the Keithleycurrent source. The maximum obtainable magnetic flux density is B = 0.78 mTwhen using the Keithley current source; this value corresponds to approximatelyH = 620 A m−1 in terms of magnetic field strength. This is sufficient to determine


Figure 6.5: Schematic drawing of the electronic setup for permalloy sensor characteriza-tion. 1: Helmholtz coils, 2: K 6221 current source, 3: sensor box, 4: temperature controlsetup, 4a: temperature control, 4b: power supply, 5: K 6221 for sensor current, 6: switchbox, 7: LabJack, 8: SR 830 lock-in amplifier, 9: K 2182A nanovoltmeter, 10: computer,11: chip with three sensors A, B, C.

the slope of the sensor response in the low-field linear region but insufficient todetermine the high field sensor response where saturation is reached. Alternatively,a Midec Type SK 150-2 is used for generating the high field values.

Temperature control

3. Sensor box. The box is made of aluminium in order to shield the experimentselectrically. In Fig. 6.6 the thermal insulation is visible and not the metal box.

4. 4a: West Instruments Model 6100+ temperature control in combination with a 4b:Kepco Model BOP 50-8 bipolar operational power supply (not visible in Fig. 6.6) toprovide the voltage for the Peltier elements. A schematic drawing of the temperaturecontrol of the sensor box is shown in Fig. 6.7, right.

Two Peltier elements and a thermocouple K-type (PE-241-14-15 and 397-1236 fromRS: www.rsonline.dk).

Controlling the actual temperature of the planar Hall sensor is accomplished bycontrolling the temperature of the environment. The temperature of the box iscontrolled externally by Peltier elements in combination with copper heat reservoirs,see Fig. 6.7 (right). The excess heat produced by the Peltier elements is removedby heat sinks. The thermocouple is buried within a copper heat reservoir thus


Figure 6.6: Experimental setup for sensor characterization. 1: Helmholtz coils, 2: K 6221current source, 3: sensor box, 4: temperature control box, 5: K 6221 for sensor current,6: switch box, 7: LabJack, 8: SR 830 lock-in amplifier, 9: K 2182A nanovoltmeter, 10:computer.

Figure 6.7: Left: Inside of sensor box (3 in Fig. 6.6). 11: chip with three sensors A, B,C, 12: socket, 13: LEMO connector, 14: thermal insulation. Right: Schematic drawing ofthe temperature control setup (connected to 4 in Fig. 6.6).


connecting the temperature control to the temperature of the copper heat reservoirs.The temperature of the copper plates is transferred to the internal of the sensor boxand, though with a time delay, the outcome is a stable temperature after equilibriumhas been reached.

Sensor current control

5. Keithley Model 6221. The sensor current is applied by a Keithley Model 6221 highprecision current source. IAC is reported as the maximum amplitude of the sinefunction generated by the current supply.

Sensor switch control

6. Switch box. In order to quickly measure the signal from all sensors, a switch boxis introduced. The switch box consists of six Reed relays Model B05-1A72-BV619(MEDER electronics AG, www.meder.com), one for each connection, that are turnedon in pairs hereby closing the electrical path across the desired sensor and connectingthe sensor to the proper input on the lock-in amplifier or the nanovoltmeter. Theswitch box introduces an additional source of noise to the setup of ∆V ≈ 2 nV butfacilitates measurements considerably.

7. LabJack Model U12. Supplies the voltage for the Reed relays in the switch box (6).

Sensor output measurement

8. Stanford Research Systems Model 830 [36]. A Stanford Research Systems Model 830lock-in amplifier is used for the noise characterization measurement and the beaddetection measurements. The reference signal for the lock-in amplifier is producedby passing the current from (5) through a low noise serial resistor of Re = 5 kΩ(RS92N/AN, www.vishay.com). The voltage drop across the resistor is sufficient forthe lock-in reference in all experiments.

The basic operation of the lock-in amplifier is governed by multiplying the detectedsignal with a reference signal and averaging over a predefined period of time, τ .The time constant defines the bandwidth as ∆f = 1/8τ for 12 dB oct−1 [36]. Thesynchronous filter is on because of the low oscillation frequency. The lock-in settingsin the various experiments are summarized in Table 6.1.

9. Keithley Model 2182A nanovoltmeter. The nanovoltmeter functions as an alternativeto the lock-in amplifier (8), mostly for trouble-shooting.

Experiment control

10. A computer controls the experiments via LabView.


Display Signal Reserve Filters Filter slopeR, θ A-B low noise sync 12 dB Oct−1

Experiment f (Hz) τ Sensitivity (µV)Noise 30 10 ms - 1 s 10

Sensitivity 330 1 s 100Temperature 330 1 s 50

Antibody 330 100 ms 100Influenza 330 100 ms 100

Table 6.1: Lock-in amplifier settings for different experiments with permalloy sensors,Batch 2. f is the applied frequency and τ is the lock-in time constant.

Sensor box

11. Chip with three planar Hall sensors A, B, C.

12. Chip holder. The chip is placed in a dedicated socket to the right inside the boxwith sensor C above B and A, respectively. Mounted below the chip is a PT100temperature sensor (www.rsonline.dk).

13. LEMO Connector Type 1B308 [37]. The electrical connections to the chip are passedthrough the box by a LEMO Connector, which has an eight connector multipole, twofor current and six for the sensor output voltages. The six voltage connections allcarry the output signal from the three sensors. Two connections per sensor to givethe voltage drop across the sensor. These six outputs are connected to the switchbox (6) used to switch between the sensors.

14. Thermal insulation.

6.3.3 Experimental conditions for permalloy sensors

This subsection describes the experimental conditions for the various experiments of Table6.1. Only conditions differing from described above are included.

Noise characterization conditions

The noise characterization measurements are performed in the setup described in Chapter5, Fig. 5.5. A Vout = 1.004 V amplitude sine output from the lock-in amplifier controls thecurrent through the sensor. A low noise resistor of Re = 5 kΩ (substituting the Re = 4.8kΩ ordinary resistor) is inserted in the series to stabilize the magnitude of current atIAC = 196 µA. The settings for the lock-in amplifier are in Table 6.1.

Sensor signal versus field

For the experiments with permalloy planar Hall sensors at high field values, B > 0.78 mT,the current through the Helmholtz coils is controlled by the Midec Type SK 150-2 power

6.4. RESULTS 81

supply. For the sensitivity measurements at low magnetic fields, the Keithley Model 6221is used.

Antibody detection conditions

In the antibody detection experiments, the experimental conditions are identical to thosedescribed above for the experimental setup. The Keithley Model 6221 current source isused to control the B-field. The settings for the lock-in amplifier are in Table 6.1. Detailson data acquisition and analysis are provided in Appendix B. The experimental resultsare presented in Chapter 7.

Immunoassay detection conditions

In the immunoassay detection experiments, the experimental conditions, data acquisition,and data analysis are identical to the antibody detection experiments described above.The experimental results are presented in Chapter 8.

6.4 Results

This and the following three sections present the experimental results obtained for the twobatches of permalloy planar Hall sensors. First the VSM measurements of the thin films.Then the electrical characterization of the fabricated sensors. The biological results arereserved for the next two chapters.

6.4.1 Vibrating sample magnetometry of the magnetic films

Fig. 6.8 shows the magnetization versus applied field experiments for the bottom-pinned(Batch 1) and top-pinned (Batch 2) stacks presented in Fig. 6.3. The left stack (Fig. 6.3)corresponds to the left graph (Fig. 6.8), and vice versa. The VSM measurements are per-formed with the applied field along the easy direction. Both measurements show successfulexchange-coupling, visualized as a shift of the hysteresis loop towards negative applied fieldstrength. The left graph reveals two loops of magnetization versus field, where the largerloop is attributed to the 20 nm thick NiFe film, and the smaller one is attributed to the5 nm thick NiFe seed layer.

The largest loop deviates from the square appearance expected for a single domain.This may be due to domain effects, which probably arise from deviation from single crys-tallinity. It seems, the MnIr film is not truly single crystalline as expected from epitaxialgrowth. It is also clear, that the seed layer, which is the smallest loop, has multiple domainbehavior. The domain structure of the seed layer is probably causing the non-epitaxialityof the pinning layer, again causing the deviation from single domain behavior of the sensinglayer. The majority of the sensing will be governed by the 20 nm NiFe layer.

In the top pinned stack, Fig. 6.8 right, similar domain behavior, again caused bydeviation from perfect crystallinity, can be observed. In this magnetic layer configurationthe anisotropy field is very low compared to the bottom pinned stack, but the larger


-30000 -20000 -10000 0 10000 20000 30000-0.20

-0.15

-0.10

-0.05

0.00

0.05

0.10

0.15

0.20

HE = 21800 A m-1

HK = 1700 A m-1

HE = 3300 A m-1

HK = 1400 A m-1

Mag

netic

mom

ent (

A m

2 )

Applied field (A m-1)

-30000 -20000 -10000 0 10000 20000 30000-0.20

-0.15

-0.10

-0.05

0.00

0.05

0.10

0.15

0.20

Mag

netic

mom

ent (

µA m

2 )

Applied field (A m-1)

HE = 3800 A m-1

HK = 200 A m-1

Figure 6.8: Vibrating sample magnetometer (VSM) measurements of magnetization ver-sus applied field strength for the two planar Hall stacks shown in Fig. 6.3. Left:substrate/Ta(30A)/NiFe(50A)/MnIr(200A)/NiFe(200A)/Ta(30A) (bottom-pinned, Batch1). Right: substrate/Ta(30A)/NiFe(200A)/MnIr(200A)/Ta(30A) (top-pinned, Batch 2).The measurements are performed with the applied field along the easy axis of the film.The left graph reveals a second loop attributed to the 5 nm NiFe seed layer.

exchange coupling compensates for that in terms of sensor sensitivity. The slight openingof the loop at negative field values may be due to impurities from the deposition process.

From the data presentation it is clear that the magnetic properties behind sensingare governed almost singularly by the exchange coupling in the case of the top pinnedstack. Whereas a considerable anisotropy contribution is present in the bottom pinnedsensor. The sum of the two contributions are comparable in the two cases and therefore thesensitivities of the two versions of planar Hall sensors will be comparable too. However, thedifference in arriving at the similar sensitivity might be important for design considerationsat a later stage of sensor development and is therefore worth mentioning.

The sensitivity, S0, depends inversely on the sum of HK and HE , and thus thesensitivity is lower for higher HK and HE . However, the linear region of the sensoris proportional to these parameters, linearity within 2 % is expected in the interval−0.23 < Hx/(HK + HE) < 0.23, see Chapter 3. With the measured HK and HE agood sensitivity is combined with an extended linear field region between H = ±900 Am−1, in flux density units B = ±1 mT.

6.5 Electrical characterization of permalloy sensors

This and the following two sections present the electrical characterization of the fabricatedexchange biased permalloy planar Hall sensors. Sensor Batch 1 was characterized atINESC-MN, and sensor Batch 2 has been characterized at MIC in the setup describedabove.

6.6. BATCH 1 83

The presented characterization of Batch 1 includes sensor resistance, sensitivity, andexperimental noise. Additionally bead detection experiments are shown. The characteri-zation of Batch 2 includes sensor resistance, noise investigation, field response, statisticalsensitivity, and temperature effects.

6.6 Batch 1

The electrical characterization measurements of Batch 1, summarized in Table 6.2, arefrom Refs. [4, 5]. The characterization was performed at INESC-MN. A Batch 1 chipwas mounted on a protoboard and a horseshoe electromagnet provided the magnetic field.Batch 1 was characterized using only DC measurements in a setup, which was not shieldedfor electronic noise.

Property Unit Theoretical Experimental Exp. std. var.RDC* Ω 25 51.9 0.5AMR % 1.9 1.3 0.02

V0 µV 458 515 10S0 µV mA−1 mT−1 38.6 30.6 2

∆V ** nV 11 250 -

Table 6.2: Theoretical values versus experimental values of the planar Hall sensors of Batch1. Last column is the experimental standard variation. (*) The DC electrical resistance,RDC, is calculated as a parallel connection of: two layers of 3 nm Ta (ρTa = 154 µΩ cm),one layer of 20 nm MnIr (ρMnIr = 175 µΩ cm), and two layers of 5 nm + 20 nm NiFe(ρNiFe = 21.7 µΩ cm) (resistivities from Gehanno et al. 1999) for a sensor geometry of10µm × 10µm. (**) The noise, ∆V , is measured during bead detection measurements ina setup without electrical shielding.

6.6.1 Bead detection, Batch 1

This subsection presents a summary of the bead detection experiments of Ref. [4]. Theaim is to verify that the exchange biased permalloy planar Hall sensor is suitable for beaddetection and only the concluding graph is presented.

First the procedure for measuring the number of beads and subsequently the signalthey produce is stated. Then the results are shown in Fig. 6.9.

Procedure

During an experiment the signal from the sensor is measured as a function of time. Experi-mental conditions are a sensor current of IDC = 10 mA, and an applied field of H = −1200A m−1. First water is added to the chip and the signal is observed to stabilize. The signalstabilization might be due to stabilization of the temperature of the chip. Water acts asa heat reservoir.


0 10 20 300

2

4

6

8

10Bead detection, Batch 1

Data Linear fit

Sig

nal (

µV)

Number of beads

Figure 6.9: Bead detection experiments from Ejsing et al. 2004. Experimental conditions:IDC = 10 mA, H = −1200 A m−1. The signal is the sensor signal for a variable numberof beads. The number of beads is estimated by visual inspection during a bead detectionexperiment. For low bead numbers, the linear fit gives 0.27(0.06) µV bead−1.

After the signal has stabilized, a solution of Micromer-M beads, diameter 2 µm, isadded to the chip using a micro pipette. The beads are flushed with water and occasionallypass by the sensor. When a cluster of beads is observed to pass the sensor, the time iswritten down along with an estimate of the number of beads in the cluster. The estimateis done by visual inspection through the microscope. For large numbers of beads, it isdifficult to judge the number due to the limited time available and due to the clusteringof beads. Hence the large error bars on the large numbers of beads.

The signal from the sensor is recorded as a function of time. When clusters of beadsare above the sensor, the signal peaks, and when the clusters leave the sensor, the signaldrops to the baseline value. The signal changes are identified with the number of beadsby matching the recorded time with the time of observing the bead clusters. Hence thesignal produced by the cluster is Vcluster = Vpeak − Vbaseline.

Results

The bead detection results are presented in Fig. 6.9. The signal as a function of thenumber of beads above the sensor is close to linear. A linear fit for the small amounts ofbeads yields an experimental single Micromer-M bead signal of

Vbead ≈ 0.27(0.06) µV

The noise level is ∆V ≈ 0.25 µV.

6.7. BATCH 2 85

Discussion

Analysis of the single bead signal from a 2 µm Micromer-M bead has shown that the sensorcurrent contributes to magnetizing the beads. Additionally, the detection of Micromer-Mbeads in zero applied field is published in Ref. [5]. In this study the beads are solelymagnetized by the DC current through the sensor. This is potentially an advantage in afuture chip design. As the external field can be excluded, there will be no need for coils,which will facilitate the fabrication and assist miniaturization.

6.7 Batch 2

Below is presented the electrical characterization measurements performed on Batch 2.Batch 2 is characterized using only AC measurements, and the setup is shielded for elec-tronic noise. The electrical characterization setup has been described above. The contri-bution to magnetization of beads by the AC current averages to zero. Thus only appliedfields should be considered, when detected by lock-in technique.

6.7.1 Investigation of noise, Batch 2

The mathematical expressions for electronic noise are given in Chapter 2. Table 6.3summarizes the estimates of noise contributions to the planar Hall sensors characterizedin this subsection.

The number of charge carriers of the Ni80Fe20 film is estimated from the volume (vol),density (%), and atomic weight (aw) of the active sensing layer. vol = 20µm×20µm×20nm,% = 8.89 g cm−3, and aw = 58.69 g mol−1. The number of charge carriers thus becomesNc = 7 · 1011 assuming one carrier per atom.

Procedure

Experimental characterization of the noise level is done at a frequency of f = 30 Hz forbandwidths set by the lock-in time constant, τ = 10 ms, τ = 30 ms, τ = 100 ms, andτ = 300 ms. The experimental noise is measured in an applied field of B = 0.5 mT inthe setup described above with a 20µm×20µm planar Hall sensor mounted. The currentthrough the sensor is set by the lock-in amplitude of VAC = 1.004 V, giving IAC = 0.196mA. The experimental noise level is estimated as the standard deviation of a series ofmeasurements with constant experimental conditions.

Results

Fig. 6.10 shows the measured experimental noise (data points) and two theoretical esti-mates (lines). The theoretical noise level for the sensor alone (red dashed line) comparedto the experimental noise is very low, including Johnson noise, shot noise, and 1/f noise.The noise from the sensor lies well below the noise from the setup (black solid line) and isthus not visible in the experimental measurements. The magnitude of experimental noisecorresponds well to the noise from the lock-in amplifier.


w (µm) t (nm) Re (Ω) γH Nc

20 20 127 0.01 7 · 1011

Johnson ( nV√Hz

) Shot ( nV√Hz

) 1/f (30 Hz) ( nV√Hz

) 1/f (330 Hz) ( nV√Hz

)1.45 2.3 2.8 0.8

Table 6.3: Electronic noise in planar Hall sensors, Batch 2. First row gives the value of theparameters for the actual sensor batch, second is the calculated noise estimates, Eq. (2.6),Eq. (2.7), and Eq. (2.8), respectively. Hooge’s constant, γH, is from Briaire et al. 1997.

0.01 0.1 1

1

10

Experimental noise Lock-in noise Sensor noise

Noi

se (n

V)

Time constant (s)

Figure 6.10: Experimental and theoretical noise of the sensor and setup, Batch 2. Thedata points are the measured noise of the setup and the solid lines are the theoretical noiseof the sensor (dashed red) and the lock-in amplifier (solid black). Experimental conditionsare B = 0.5 mT, f = 30 Hz, and IAC = 0.196 mA. The theoretical sensor noise is thesum of the Johnson, shot noise, and the 1/f noise for a 20µm×20µm planar Hall sensor.The graph is a log-log plot revealing the inverse square root dependence of the theoreticalnoise level.

6.7. BATCH 2 87

The noise spectrum of the setup is inspected with a Spectrum Analyzer (HewlettPackard Model 4396A, Network/Spectrum Analyzer). The outcome of that experimentproves that f = 30 Hz and f = 330 Hz are good choices of measurement frequencies asthey are situated at relatively quiet positions in the noise spectrum.

Assay noise

Increasing the time constant on the lock-in amplifier decreases the expected noise, andincreasing the frequency to f = 330 Hz decreases the 1/f noise level to ∆V/

√∆f = 0.8

nV/√

Hz. For integration times of τ = 100 ms (∆f = 1.25 Hz) and averaging over 100 datapoints the theoretical noise level of one experiment becomes: Ntot ·

√∆f/

√(#meas− 1) ≈

0.3 nV, where Ntot is the sum of Johnson noise, shot noise, and 1/f noise added usingthe law of error propagation. An experimental standard error of ∆V ≈ 1 nV is obtained,which is the background for the bead detection limit estimate in Chapter 4.

6.7.2 Sensor response vs. field, Batch 2

The characterization of the field response of permalloy planar Hall sensors are presentedin Fig. 6.11, and Fig. 6.12. For ease of comparison to theoretical calculations, the valuemeasured at 30 oC is chosen for all presented data though a temperature interval between15 oC and 45 oC has been investigated.

The planar Hall signal’s behavior up to high field values is shown in Fig. 6.11. Thegenerated curve is Eq. (3.9) with V0 = I AMR ρavt

−1 = 223.5 µV and HE +HK estimatedfrom the hysteresis measurements, Fig. 6.8 (right). From the experimental data an off-setof V = 3.45 µV has been subtracted before comparing to the theoretical curve. Goodagreement is observed at low applied fields up to approximately B = 2 mT, where thecurves start deviating.

The insert of Fig. 6.11 presents a fit to the experimental data. The best fit yieldsparameters: HE = 3458 A m−1, HK = 1.86 A m−1, and V0 = 191.4 µV. These areclose to the expected values. Expected values: HE = 3800 A m−1, HK = 200 A m−1

obtained from the VSM measurements (Fig. 6.8), and V0 = 223.5 µV. However, the VSMmeasurements will most likely over-estimate the value of HK , which is small compared tothe experimental accuracy.

The sensor response as a function of applied magnetic field yields a straight line withthe slope S0 = 44.8(0.2) µV mT−1 mA−1 at T = 30 oC. This value is the average valueof the sensitivity of 31 sensors, presented in Fig. 6.12. Fig. 6.12 shows the sensitivityfor different planar Hall sensors of equal design (20µm × 20µm sensitive areas). Thelarger error bars associated with the second half of the measurements are due to less datapoints used to determine the slope. The reason is to save time on the measurements.The deviation between the highest and the lowest sensitivity is approximately 10 %. Thedotted line shows the theoretically expected value given by Eq. (3.19), which is S0 = 44.3µV mT−1 mA−1.


0 2 4 6 8 10 12 140

20

40

60

80

100

High field responseV

(µV

)

Applied field (mT)

Generated (no fitting) Data

Fit to data

Figure 6.11: Experimental and theoretical (generated from VSM measurements of HK andHE) high field behavior of the permalloy planar Hall sensor. In the experimental data, anoff-set of 3.45 µV has been subtracted. The insert shows a fit of the combined Eq. (3.9)and Eq. (3.16) to the experimental data. This yields HE = 3458 A m−1, HK = 1.86 Am−1, and V0 = 191.4 µV.

0 5 10 15 20 25 30 3540

41

42

43

44

45

46

47

48

49

50 20 µm sensors

Sen

sitiv

ity (

V m

T-1 m

A-1)

Sensor

Figure 6.12: Experimental sensitivities for 31 planar Hall sensors of 20µm×20µm sensitiveareas. The dotted line is the theoretical value.

6.7. BATCH 2 89

6.7.3 Investigation of temperature dependence, Batch 2

Due to fluctuations in room temperature, the sensor signal is observed to vary with tem-perature. When the temperature rises, the signal rises, and when the temperature drops,the signal drops, see Fig. 6.13 and Fig. 6.14.

Procedure

A constant field of B = 0.54 mT is applied during the measurements. The temperatureof the copper plates is set at T = 20 oC through T = 40 oC and back with step sizes of∆T = 5 oC. The actual temperature inside the box is measured with the Pt100 elementand the values are reported on the graph, Fig. 6.13. Each value corresponds to the endtime at each temperature interval, which is the most stable value and approaches anequilibrium value. Evidently the measured temperatures vary more than is attempted bysetting the temperature of the copper plates.

Results

Fig. 6.13 shows a nine hour time trace of the sensor signal, in which the temperature ischanged every hour. There is a clear stepwise dependence when altering the temperature.One can observe that the signal comes to rest when the temperature equilibrium is reached,the steps flatten, and the temperature of T = 30.9 oC, setpoint T = 30 oC, is the moststable operating temperature of the ones tested. At this setpoint, the power supply forthe Peltier elements has less work load in maintaining a constant temperature due to thevicinity of room temperature.

The results are summarized in Fig. 6.14 (left), where the signal is plotted as a functionof temperature. The path up differs slightly from the path down showing a tendency ofhysteresis. However, it is difficult to determine the actual signal at the time where thetemperature measurement is taken due to a much higher noise level than usual. This extranoise probably arises because of temperature fluctuations when adjusting the temperature.Rather large errors to the data points result. A linear curve is fitted to the data points toobtain an estimate of the magnitude of the signal temperature dependence. ∆V ≈ 55 nVoC−1 corresponding to a change of approximately 0.2 % oC−1.

Additionally, the sensitivity, S0, and the electrical resistivity, Re, are measured as afunction of temperature. These are plotted in Fig. 6.14 (right) along with V . S0 changesapproximately 0.24 % oC−1. The change in Re as a function of temperature is sufficientto account for the changes in V and S0 within experimental errors.

Discussion

These temperature variations are comparable to those of Ky’s experimental data [30].Ky gets a temperature variation of thin nickel films in the order of 0.2-0.25 % oC−1, seeChapter 2. Since the permalloy films are composed of 80 % nickel, their behavior is likelyto be similar. Montaigne et. al. reports approximately 0.3 % oC−1 change in Re for theirpermalloy films [1].


0 1 2 3 4 5 6 7 8 927.0

27.5

28.0

28.5

29.0

29.5

15.10

21.80

30.60

38.20

45.20

38.20

30.90

22.10

Temperature time trace

Signa

l (µV

)

Time (h)

15.30

Figure 6.13: Temperature dependence of the sensor signal. The graph shows a nine hourtime trace of the sensor signal, where the temperature has been changed every hour. Themeasured temperature values are included on the graph.

15 20 25 30 35 40 45

27.5

28.0

28.5

29.0

29.5

Temperature dependence

Signa

l (µV

)

Temperature (0C)

10 15 20 25 30 35 40 45 50

0.93

0.94

0.95

0.96

0.97

0.98

0.99

1.00

1.01

Signal change compared to resistance change

V/Vmax

S/Smax

R/Rmax

Nor

malized

value

s

Temperature (0C)

Figure 6.14: Temperature dependence of the sensor signal. Left: sensor signal as a functionof temperature. A linear fit reveals the change as 55 nV oC−1 corresponding to 0.2 % oC−1

change in signal. Right: signal magnitude, V , and sensitivity, S0, of the sensor is comparedto the change in resistance, Re. V , S, and Re are normalized to their value at the highesttemperature.

6.7. BATCH 2 91

The experiment is repeated with no field applied in order to observe the off-set as afunction of temperature. This results in the same temperature dependence. Hence themagnetic field dependent contributions, anything apart from the resistivity, can be ruledout as they depend on the magnetic state of the film.

A possible approach to controlling these effects would be to keep the voltage constantinstead of the current. If only the electrical resistance changes, the sensor output voltagewould be constant for a constant applied voltage.

6.7.4 Summary, Batch 2

All the above investigations of Batch 2 sensors are summarized in Table 6.4. Generallygood agreement is observed between expected or calculated values and experimentallyobserved values. In the experimental value for S0, the current is assumed to flow solelythrough the permalloy layer of the stack. Shunting of the current through the other layersof the stack can be simulated using the commercially available package FEMlab, this yieldsV0 = 211 µV and S0 = 49.3 µV mA−1 mT−1 for the experimental values.

Property Unit Expected Measured Exp. std. var.ρstack* µΩ cm 42.8 46.8 0.5AMR % 1.9 2.06 0.02HE** A m−1 3800 (VSM) 3500 (fit) -HK** A m−1 200 (VSM) 2 (fit) -

V0 µV 223.5 190 10S0 µV mA−1 mT−1 44.3 44.8 0.2

Noise*** nV 0.9 3.9 0.5Assay noise nV 0.3 1 -Temperature 0.2-0.25 (Ky) 0.2 (V ) 0.01 (V )

drift % oC−1 0.3 (Montaigne) 0.24 (S0) 0.02 (S0)

Table 6.4: Calculated/expected values versus measured values of the planar Hall sensorsof Batch 2. Last column is the experimental standard variation on the measured value.(*) ρstack is calculated as a parallel connection of two 3 nm thick Ta layers (ρTa = 154 µΩcm), 20 nm MnIr (ρMnIr = 175 µΩ cm), and 20 nm NiFe (ρNiFe = 21.7 µΩ cm) (resistivitiesfrom Gehanno et al. 1999) for a sensor geometry of 20µm× 20µm. (**) The values givenas the ”measured” HK and HE are obtained from the fitted curve to the experimentaldata, Fig. 6.11, whereas ”expected” are from the VSM measurements, Fig. 6.8 (right).(***) The bandwidth is set by the time constant τ = 1 s, and the temperature is T = 30oC. The expected noise value is the sum of the Johnson noise, shot noise, and the 1/fnoise for the sensor alone.


6.8 Discussion of S/N improvement

This section discusses briefly some possibilities of improving S/N in the electrical setupused for characterizing the Batch 2 sensors.

The noise level of the planar Hall sensors is independent of the applied magnetic fluxdensity. In the sensor Batch 2, the sensor response starts deviating from linear with respectto the applied field at approximately B = 1 mT, which puts a natural limit to the fieldfor magnetizing the beads. Furthermore, currents of up to I = 10 mA can be appliedwithout destroying the sensor, and the tests are done with much less current than that.Due to practical circumstances only IAC = 1 mA (maximum amplitude) is used. Pushingthe limit of field and current, a factor of 10-20 could be gained.

The noise from the planar Hall sensors is dominated by the noise from the lock-inamplifier, see Fig. 6.10. The introduction of a low-noise pre-amplifier between the sensorand the lock-in amplifier would help to utilize the full scale of the planar Hall sensor.An estimate of the signal-to-noise enhancement following this procedure is approximately5. In total it should be possible to enhance S/N 50-100 times, which makes detection ofsingle 250 nm beads within reach of the actual sensor batch. Table 4.3 predicts that 62Nanomag-D beads on a sensor will yield S/N = 2.

6.9 Conclusion

This chapter has presented two approaches of exchange biasing permalloy planar Hallsensors. Batch 1 constitutes a bottom-pinned approach and Batch 2 a top-pinned. Thefabrication procedure of the two sensor batches has been described in detail.

Batch 1 was fabricated and characterized completely during the author’s stay atINESC-MN, Portugal. The characterization was only DC, and no electrical shielding ofnoise was conducted during the measurements. Batch 2 has been fabricated essentially atMIC, and efforts are put into electrical shielding and noise reduction in the experimentalsetup described in the text.

The VSM measurements show that a successful exchange coupling is induced in thethin magnetic films. Thus the zero field starting point of the film’s magnetization is, andwill be, well-defined.

For Batch 1, electrical characterization measurements have been presented and com-pared to theory. The comparison is reasonable except from RDC. Additionally, beaddetection experiments obtained in [4] are included. The single bead signal reported froma Micromer-M bead is 0.27 µV, though with a sensor noise level in the same order ofmagnitude [4]. Using DC sensor currents, Micromer-M beads can be detected in zeroapplied field [5], which will facilitate the electronics in an eventual point-of-care device.No biological experiments have been performed with Batch 1 sensors.

For Batch 2, sensor noise, sensor field response, and temperature dependence have beenthoroughly investigated. The experimental values compare well with theory, see Table 6.4.Finally, it is estimated that S/N could be improved 50-100 times in the present setup byadjusting field, current, and introducing a pre-amplifier. Since the calculated detection

6.9. CONCLUSION 93

limit is 62 Nanomag-D beads, single bead detection should be within reach for Batch 2sensors.

Chapter 7

Antibody detection

In this chapter, detection of antibodies immobilized on top of the planar Hall sensorsis realized. Batch 2 sensors are used. Biotinylated monoclonal antibodies in varyingconcentrations are immobilized on top of a sensor and streptavidin coated beads bindspecifically to the biotin label. The presence of antibodies is detected by detecting thebeads.

As the surface where proteins are immobilized, SU-8 photoresist is chosen. SU-8 is anegative tone photosensitive polymer, and can thus be patterned directly by lithographydescribed previously. The polymer is non-conducting and shunting of the sensor current isavoided. Furthermore, SU-8 can be spun on the wafer in a very thin layer, 0.5 µm, whichis an advantage since the sensitivity of the sensor towards the beads strongly depends ondistance.

7.1 Immobilizing antibodies

In order to use the planar Hall sensor in an immunoassay, the appropriate antibodies mustbe immobilized on the surface. When discovering that the photosensitive polymer SU-8has a high binding capacity for biomolecules, this property was immediately pursued fordetection of DNA samples [34]. During this work, it was discovered that SU-8 has a highbinding capacity not only to DNA fragments but also to proteins. This is the motivationfor choosing SU-8 for attaching antibodies on top of the planar Hall sensors though thesystem is new and the binding mechanism so far not known.

This section presents the immobilization procedure. First coating and structuring ofSU-8 is addressed, then the binding of antibodies to the SU-8 surface.

7.1.1 SU-8

The negative tone photosensitive polymer SU-8 can be purchased at MicroChem (www.micro-resist.de) and both the first SU-8 series and the SU-8 2000 series can be used to attachDNA and proteins. The only difference is the solvent on which the solution is based. Forthe coating of planar Hall sensors, diluted SU-8 2002 is used.

95

96 CHAPTER 7. ANTIBODY DETECTION

During experiments with DNA [34] a very high autofluorescence from SU-8 has beenobserved. This calls for using a very thin layer of SU-8 for the fluorescence detection.Additionally, the planar Hall sensors are sensitive to the sensor-bead separation, whichalso implies the need for thin SU-8 layers. When SU-8 2002 is mixed in a 1:1 weight relationwith cyclopentanone, the solvent for the 2000 series, a thickness of approximately 0.5 µmis obtained at a spinning speed of 2000 rpm. This reduces the fluorescence backgroundsignal considerably, and the spacing between sensor and beads is theoretically tolerable,see Fig. 4.5. Today, this solid content is available directly as the product SU-8 2000.5.

Processing of the SU-8 is simply following the instructions given by the supplier, thoughwith minor modifications. An important condition for SU-8 to stick tightly to the surfaceis a dry and clean surface. The recommended procedure to ensure this is to dehydratethe wafer before coating at T = 250 oC for at least four hours. Usually the wafer is leftin the oven over night and ready for coating the following morning. This procedure isnot compatible with magnetic sensor technology. The magnetic properties of the film aredestroyed at temperatures above T = 100-150 oC. Instead of using the 250 oC oven forsensor Batch 2, the dehydration is done at T = 100 oC for one hour in a vacuum chamber.

The SU-8 2000.5 yields a layer thickness of approximately 0.5 µm at a spinning speed of2000 rpm. After coating a prebake is performed directly on the spinner using the followingprocedure:

1. The wafer is placed at a 5 mm distance above the spinner hotplate, which is set atT = 90 oC. Here it bakes for 30 s.

2. Then the wafer is moved to 2 mm distance, where it bakes for 30 s.

3. Then it is placed directly at the hotplate for 2 min.

4. The two preliminary steps are repeated backwards (2 and 1), and the wafer is trans-ferred out of the spinner hotplate.

The stepped sequence is recommended by the supplier in order to avoid rapid heating ofthe SU-8 when baking out the solvent. Too rapid heating or cooling can result in cracks inthe SU-8 layer. After baking, the wafer is exposed through a mask at the desired locationsfor cross linking, see Fig. 6.4 fifth step. The light is UV and the power continuously 9 mWcm−2 (setting: Cl2 on the KS-aligner), the exposure lasts 30 s. Cross linking is hereaftercompleted in a postbake step, which is identical to the prebake sequence just described,and the wafer is developed for 1 min in 1-methoxy-2-propyl acetate (PGMEA) and rinsedin isopropanol (IPA). The procedure is summarized in Table 7.1.

Dehydration Coating Pre-bake Exposure Post-bake Development1 h, 100 oC (vac.) 2000 rpm 1-4 30 s (Cl2) 1-4 1 min (PGMEA)

Table 7.1: SU-8 process for 0.5 µm.

7.1. IMMOBILIZING ANTIBODIES 97

7.1.2 Antibody binding to SU-8

This subsection describes how antibodies are bound to the SU-8 surface. The antibodiesare biotinylated monoclonal influenza antibodies [38] provided by Statens Serum Institut(www.ssi.dk). Two detection systems are used, fluorescent magnetic beads coated withstreptavidin, and Cy-3 conjugated streptavidin. The fluorescent signal is obtained byscanning the microscope slide in an Applied Precision Model ArrayWorx laser scanningsystem.

First the properties of the fluorescent magnetic beads are described, then the immo-bilization procedure, the results, and the fluorescent reference system is presented.

Beads

For the biotin-streptavidin reference system, two commercial bead products are used. TheSpheroTM beads are fluorescent as well as magnetic, whereas the Nanomag-D beads areonly magnetic. The properties of the beads are summarized in Table 7.2. χm of theSpheroTM beads has not been measured on the VSM, but their χm is small compared tothe Nanomag-D beads.

Bead Diameter Surface Fluorescence χm

SpheroTM Nile Red 360 nm Streptavidin 0.2-0.39 µm smallNanomag-D 250 nm Streptavidin none 3.2(0.1) (SI)*

Table 7.2: Physical properties of magnetic beads (suppliers: www.spherotech.com andwww.micromod.de). (*) χm for the Nanomag-D beads is measured on the VSM, seeChapter 4.

Procedure

Monoclonal antibodies are dissolved in PBS (phosphorous buffered saline) in concentra-tions c = 0.5, 1.0, 5.0, 10, 50, 100, 500, and 5000 µg ml−1. Hereafter 0.5 µl of the variousantibody concentrations are placed on the SU-8 surface using a micro pipette. The sampleis left at T = 37 C for one hour in a humid atmosphere. During this time period theantibodies adsorb on the SU-8 surface, where they are immobilized.

Washing in PBS removes the unbound antibodies, and blocking the rest of the surfaceis done in a solution of c = 5 mg ml−1 BSA (bovine serum albumin) in PBS with additionof 5 % fat free milk. Other blocking systems, such as BSA, rabbit serum, and bovineserum have been tested, but the most promising results in terms of minimizing unspecificbinding are found for the BSA + milk system. Then the slide is washed 5 minutes in PBSand rinsed briefly in DI water.

Streptavidin coated fluorescent beads, SpheroTM beads, are placed on the microscopeslide and left for half an hour at room temperature. The beads bind to the antibodies onthe SU-8 surface via the biotin-streptavidin bond.


1 10 100 1000 10000

0

20

40

60

80

100

Biotin-streptavidinsaturation curve for fluorescent beads

Fluorescent signal/background

S/B

G

Antibody concentration (µg/ml)

Figure 7.1: Saturation curve for the biotin-streptavidin system of SpheroTM beads, flu-orescent and magnetic. S: signal, BG: background. Approximate linearity of signal toconcentration is found for protein concentrations between c = 10 µg ml−1 and c = 1000µg ml−1.

1 10 100 1000 10000-20

0

20

40

60

80

100

120

140

160

Biotin-streptavidinsaturation curve for Cy3-conjugated streptavidin

Fluorescent signal/background

S/B

G


Figure 7.2: Saturation curve for the biotin-streptavidin system with Cy3-conjugated strep-tavidin. S: signal, BG: background.

7.2. ANTIBODY BINDING REALIZED USING THE PLANAR HALL SENSOR 99

Figure 7.3: Procedure for measuring biotinylated monoclonal antibodies (Mab-biotin).The magnetic sensor measures the presence of the streptavidin coated magnetic bead,which is bound to the sensor by the biotinylated antibodies.

Results

The signal to background curve obtained by scanning the fluorescent signal from thebeads is presented in Fig. 7.1. The signal, S, is measured as the average intensity ofthe fluorescent spots minus the background intensity, and BG is the average intensityof the background. It is clear that protein concentrations below c = 10 µg ml−1 areundetectable, and that for protein concentrations above c = 1000 µg ml−1 saturationis reached. Between these two antibody concentrations an approximately linear relationbetween signal and concentration is obtained.

Reference system

An identical experiment is performed using Cy3-conjugated streptavidin as detection sys-tem instead of streptavidin coated fluorescent beads (Fig. 7.2). For this system a proteinconcentration of c = 5 µg ml−1 is the lower detection limit, and saturation is reached atapproximately c = 100 µg ml−1. For most concentrations the signal-to-noise is approxi-mately 50 % higher than that of the fluorescent bead detection system.

7.2 Antibody binding realized using the planar Hall sensor

This section describes the detection of monoclonal antibodies by the planar Hall sensor.First the chemical binding procedure is described, then how to obtain the resulting signalfrom the beads. Results are presented without and with subtraction of the reference.

7.2.1 Procedure

A drawing of the procedure for detection of monoclonal influenza antibodies using theplanar Hall sensor is shown in Fig. 7.3. Below the procedure steps are described.

1. The signal from the sensor and a reference sensor is measured. A correlated drift isobserved, which disappears when the reference signal is subtracted from the sensorsignal. See Fig. 7.4 and Fig. 7.5.


0 100 200 300 400 500 600-0.10

-0.08

-0.06

-0.04

-0.02

0.00

0.02

0.04

Chip 201, Sensor drift (30 C)

A B C

Sig

nal-m

ean

(µV

)

Time (s)

Figure 7.4: Initial signal from three sensors, VA, VB, VC, as a function of time. Measuredbefore processing and without beads. The mean value has been subtracted in order tovisualize the drift, which on average is in the order of 0.2-0.5 % of the total signal. Onemeasurement is the average of 100 points and lasts 10 s. Experimental conditions areB = 0.5 mT, IAC = 1 mA (K6221), T = 30 C.

0 200 400 600 800 100026.38

26.40

26.42

26.44

26.46

26.48 Chip 201, Sensor A-B (30 C)

03-11-2005 02-11-2005 Demount

Sig

nal-r

efer

ence

(µV

)

Time (s)

Figure 7.5: Signal from sensor A where the reference sensor B has been subtracted, VA−VB.Measured before processing and without beads. Time and demounting the sensor (takethe chip out of the socket and put it back in) leave the response unchanged. Experimentalconditions are B = 0.5 mT, IAC = 1 mA (K6221), T = 30 C.


2. Monoclonal antibodies are immobilized on top of the sensors. The concentration isvaried between c = 0 µg ml−1 and c = 100 µg ml−1 in PBS. The antibodies used arelabelled with biotin. Procedure: 1 hour at T = 37 C in a humid atmosphere.

3. The remaining SU-8 surface is blocked with BSA and milk. Blocking the surface witha protein solution ensures that the detector antibody cannot attach to the surface.It can only attach to the antigen, thus decreasing the background signal. Procedure:Half an hour at room temperature. Blocking solution: cBSA = 5 mg ml−1 in PBS +5 % fat free milk.

4. Streptavidin coated magnetic beads are washed three times in PBS. A bead concen-tration of c = 1 mg ml−1 is used. The bead solution is placed on the chip for thestreptavidin-biotin binding to take place. Half an hour at room temperature with”stirring” on a micro titer.

5. The unbound beads are washed off in PBS (5 min. + stirring), and the chip is leftovernight to dry.

6. The signal from the sensor is measured, the reference signal is subtracted, andcomparison with the previous signal gives the signal from the beads, Vbeads =Vafter − Vbefore.

Procedure details

Before measuring care must be taken not to damage the chip during chemical processing.The wire bonds are very fragile and can be damaged by chemicals, rough or carelesshandling, air flow from the nitrogen valve, or by electrical pulses in the system. Toprovide protection against handling, the wire bonds are covered in silicon gel (ElastosilE 41, Wacker, Germany). This protects the chip against most physical and chemicalhandling. However, electrical pulses must be avoided in order not to blast the wire bondsor the leads on the chip.

Another issue is that no beads must come into contact with the reference sensor.Only a few beads can be detected by the sensor and in the case where the reference iscontaminated a false baseline would be subtracted. Contamination is avoided using a fatbarrier on top of the reference sensor that repels liquids and thus solutions with beads.

One measurement is the average of the sensor signal from 100 observations. Theduration of 100 samples is approximately 10 s. Then the switch box switches to the nextsensor and the measurement of another 100 samples is performed. The measurementsequence is sensor A, sensor B, sensor C, and then back to sensor A, and the sequenceis run a prespecified number of times. The time delay between the actual biochemistrysensor and the reference sensor is thus 10 s plus the switching time, which is 100 ms.This means that subtracting the reference is not done simultaneously as in differentialmeasurements.

This is the reason for controlling the temperature of the experiment. After stabi-lization, the temperature variations are small compared to the electrical noise, and the


reference can be used to correct for an eventual change in off-set from demounting and re-mounting the sample. That the procedure is accurate can be seen from the investigationspresented in Appendix B.

When the signal without beads has been measured, the biotinylated monoclonal anti-bodies are immobilized on top of sensor A. Sensor B has a fat barrier on top, and sensorC is left without antibodies to measure the background signal from unspecifically boundbeads. The immobilization procedure is described above for the fluorescence detectionmeasurements along with blocking of the residual SU-8 surface.

After immobilization of antibodies, the chip surface is carefully dried in a nitrogen flow.The surface should be more or less dry for the bead solution to stay where it is placedon top of the chip. 1-2 µl of streptavidin coated beads are placed on top of sensor Aand sensor C, where they are left for half an hour to react with the biotin on the surface.Stirring of the chip during the reaction ensures that beads come into contact with thebiotinylated antibodies and minimizes gravitational settlement at undesired places.

After the biotin-streptavidin bond is formed the chip is washed in order to removeunspecifically bound beads. It is however impossible to remove all beads. They have atendency to get stuck everywhere on the chip. Despite all efforts of removing them with amagnet or putting soap in the buffer there still remain beads where they are not supposedto be. The signal from zero antibody concentration corresponds to the background signalfrom the unspecifically bound beads.

Water on the chip will have an effect on the temperature of the chip. Therefore itis essential that the chip is completely dry before measuring the signal from the beads.The chip is dried with a nitrogen flow and left overnight, which ensures a completely drychip. After remounting in the socket the chip is left for an hour with applied current andtemperature control before starting a measurement. This way the temperature of the chipas well as the electrical offset will be stable during the measurement.

With beads on top of the sensor, the signal stabilizes slowly after the applied field isswitched from B = +0.5 mT to B = −0.5 mT, see Fig. 7.6. Hence after switching thefield, the signal stabilizes for 10 s before performing the electrical measurement.

This observation is not due to the lock-in amplifier or the switch box. It only happenswhen the applied field is changed, not when the switch box changes back to the sensor.Additionally, everything is left for an hour to stabilize, which has also been verified ex-perimentally to be sufficient for stabilization (data not shown). The only thing that ischanged is the applied magnetic field, hence it must be a reaction to a changed magneticfield. The speculation is, that what is seen is due to domain effects in the permalloy film,but no proof for that has been found.

7.2.2 Results

The measured signals are normalized to the sensor sensitivity, applied current and theapplied flux density, B (measured in mT), as shown in Eq. (7.1). This yields a number, β,which is independent on sensor specific parameters, and only depends on the field sensed


7 8 9 10 11 12 13 1440.70

40.75

40.80

40.85

40.90

40.95

41.00

41.05 Switch B (+0.5mT to -0.5mT)

Sig

nal (

µV)

Time (s)

Figure 7.6: Stabilization time for a planar Hall sensor with beads. The applied field isswitched from B = +0.5 mT to B = −0.5 mT at time = 7.0 s.

by the sensor.

β =signal

sensitivity · sense current ·B(mT)(7.1)

Taking the mean of the values at B = +0.5 mT and B = −0.5 mT subtracts the offset.

∆β =12β(+0.5 mT) +

12β(−0.5 mT) (7.2)

The reference, sensor B, is then subtracted from the sample, e.g. sensor A.

∆βA −∆βB (7.3)

The reference does not change when beads are added on top of sensor A, and thus Eq. (7.4)gives the signal from the beads normalized to B = 1 mT applied flux density.

∆βbeads = (∆βA −∆βB)after − (∆βA −∆βB)before = ∆βafter −∆βbefore (7.4)

For a detailed description of data acquisition and analysis, see Appendix B. In the mea-surements, the normalization is to B = ±0.5 mT, but taking the mean value correspondsto a magnetic flux span of 1 mT. Hence, the normalization will be referred to as B = 1mT.


Without reference

The first measurements of antibody detection are performed without subtracting the ref-erence sensor, and the results are shown in Fig. 7.7. A single measurement point seems notto follow the trend of increasing signal for increasing antibody concentration. However,visual inspection of this sensor in the microscope shows that only a few beads are presenton top of the sensor. Not enough beads have come into contact with the sample or thechemical reaction is not sufficient in this case.

Table 7.3 compares the data points of Fig. 7.7 to the fluorescent measurements ofFig. 7.1 and Fig. 7.2. The signal-to-noise, S/N , is calculated as the signal from theNanomag-D beads divided by the error related to the measurement. Compared withfluorescent signal divided by background, S/BG, the magnetic signal is much lower for thehighest antibody concentration. However, for low antibody concentrations the magneticsignal is more promising than the fluorescent signal.

With reference, experiment 1

Detection of monoclonal antibodies with a reference sensor subtracted is shown in Fig. 7.8.

The figure shows the signal from the beads as a function of antibody concentration.The value ∆βbeads = ∆βafter −∆βbefore reflects the magnetic flux produced by the beadsat the sensor surface in an applied magnetic flux density of B = 1 mT, and the highestvalue is ∆βbeads = 0.011.

Two observations are extracted from the figure. First, the signal rises with increasingantibody concentration. Second, a rather large signal arises from zero concentration. Thisis a problem because it reduces the sensitivity of the bead assay.

Fig. 7.9 shows pictures taken after the immobilization of streptavidin coated beads.Left: antibody concentration c = 50 µg ml−1. Right: antibody concentration c = 5 µgml−1. The bead coverage is larger in Fig. 7.9 (left) than in (right) as expected since theantibody concentration in higher.

The pictures show that more beads are bound at sites outside the SU-8 pad covering thesensor than on the SU-8 pad itself. This is unfortunate since the aim is to have antibodiesonly on top of the sensor. There is a residual monolayer of SU-8 left after development,and it seems just as good as the thicker layer when it comes to binding antibodies.

With reference, experiment 2

Repeating the measurements yields the response shown in Fig. 7.10. Again an increasingsignal as a function of increasing antibody concentration is obtained. The slope of thecurve is similar to the slope in Fig. 7.8 but the signal obtained from the zero concentrationpoint has decreased. This is fortunate because the sensitivity depends on the magnitudeof unspecific binding.

Pictures of the sensors with different antibody concentrations are presented in Fig. 7.11.The respective antibody concentrations are: top left: c = 100 µg ml−1, top right: c = 50µg ml−1, bottom left: c = 10 µg ml−1, bottom right: c = 0 µg ml−1. The bead coverage


10 100

100

200

300

No reference

Sig

nal c

hang

e (n

V)


Figure 7.7: Detection of monoclonal antibodies using the planar Hall sensor. These mea-surements are performed without subtracting the reference sensor from the sensor withantibodies.

Antibody concentration 5 µg ml−1 10 µg ml−1 50 µg ml−1 100 µg ml−1

S/N Magnetic beads 4.5 31 5 34S/BG Cy3-streptavidin 1.3 37 136 140S/BG Fluorescent beads 5.3 6.7 41 49

Table 7.3: Comparison between magnetic and fluorescent detection systems. The mag-netic beads are Nanomag-D beads, 250 nm in diameter. The signal is obtained withoutreference subtraction, data from Fig. 7.7. S/N is the signal from the beads divided by thestatistical error, S/BG is the fluorescent signal intensity minus the background dividedby the background intensity.


0 20 40 60 80 100

0.005

0.006

0.007

0.008

0.009

0.010

0.011

Reference subtracted

afte

r-be

fore

Antibody concentration ( g ml-1)

Figure 7.8: Detection of monoclonal antibodies realized using the planar Hall sensor. Thesignal from the beads, ∆βbeads = ∆βafter−∆βbefore, is normalized to the applied field. Fordetails on data acquisition and analysis see Appendix B.

Figure 7.9: Pictures of the bead coverage for antibody concentrations of c = 50 µgml−1(left) and c = 5 µg ml−1(right). These correspond to the respective electrical mea-surements of Fig. 7.8.


-10 0 10 20 30 40 50 60 70 80 90 100 110

0.000

0.001

0.002

0.003

0.004

0.005

0.006

0.007

Reference subtractedaf

ter-

befo

re


Figure 7.10: Detection of monoclonal antibodies realized using the planar Hall sensor. Thesignal from the beads, ∆βbeads = ∆βafter−∆βbefore, is normalized to the applied field. Fordetails on data acquisition and analysis see Appendix B.

Figure 7.11: Pictures of the bead coverage corresponding to different antibody concentra-tions. Decreasing antibody concentration from left to right. Top left: c = 100 µg ml−1,top right: c = 50 µg ml−1, bottom left: c = 10 µg ml−1, bottom right: c = 0 µg ml−1.


Experiment 1 cmab (µg ml−1) ξoutside ξsensor βξ ∆βbeads (Fig. 7.8)Fig. 7.9 (left) 50 0.75 0.50 0.008 0.0086Fig. 7.9 (right) 5 0.25 0.10 0.004 0.0068

Table 7.4: Estimates of layer coverage, ξ, obtained from the pictures of the sensors withbeads, Fig. 7.9. The corresponding estimates of the field produced, βξ (normalized to theapplied field), are calculated using Eq. (4.33). The ∆βbeads values are obtained experi-mentally, Fig. 7.8.

decreases with decreasing antibody concentration. Though the picture in the bottom leftcorner at a first glimpse shows a higher bead concentration than its predecessor, dirt isobstructing the view of the beads, and the true coverage is probably less than observed inthe picture.

7.3 Discussion

The positive signal in Fig. 7.8 and Fig. 7.10 has been addressed in Chapter 4, Eq. (4.33).

β = 0.0257 · ξoutside − 0.0225 · ξsensor

where ξ is the bead coverage with respect to a closely packed monolayer. The equationassumes non-interacting beads. If rough estimates on the bead coverage are obtainedby visual inspection of Fig. 7.9, the field produced by the respective layer coverage canbe estimated using Eq. (4.33). The resulting estimates are presented in Table 7.4 alongwith the experimentally determined ∆βbeads-values from Fig. 7.8. The estimates comparereasonably well with the data, at least the order of magnitude is in place.

The maximum value obtained in the experiments is β = 0.011. This correspondsroughly to an estimated difference in layer coverage of one half for the approximation toEq. (4.33), β ≈ 0.024(ξoutside − ξsensor). It is fair to assume that the layer coverage is lessthan one. A closely packed coverage of beads cannot be expected due to steric hindrancewhen the streptavidin on the beads reacts with the biotin on the surface. It is also fair toassume that the layer coverage on top of the sensor is less than outside. The beads haveto jump a step of 0.5 µm, twice their diameter, to bind to the biotin immobilized here.

The zero points in the figures reflect the unspecific binding in the system. The unspe-cific binding is considerably reduced in Fig. 7.10 compared to Fig. 7.8. The reason for thedecrease in unspecific binding is probably a slight change in the washing procedure. Thetweezers holding the chip in the PBS buffer are tapped gently on the rim of the glass. Themechanical kick might have loosened the unspecifically bound beads. Another explanationcould be that the streptavidin is older and thus binds less tightly to biotin and perhapsother proteins. Reducing the unspecific binding will be an issue of great importance infuture work with this assay.

A general observation extracted from Fig. 7.8 and Fig. 7.10 is that the signal increasesfor increasing antibody concentration. This is exactly the observation which is desiredfrom this study. The more beads attached to the surface, the higher the signal from

7.4. CONCLUSION 109

the beads. This finding indicates that a negative control in an immunoassay will producelower signal than the sample assuming that the sample binds more beads than the negativecontrol.

The images, Fig. 7.9 and Fig. 7.11, show where the beads are attached after incubation.There is a general tendency of catching some beads at the edges of the SU-8 pad. Thereis however a more striking tendency of bead capture at the edges of the voltage leads.These leads are made of the same material as the sensor, i.e. the pinned ferromagneticstack shown in Fig. 6.3. The easy axis is horizontal with respect to the photos and thusthe only place where the fringing field has a gradient is at the edges of the voltage leads,vertically in the photos. These arguments suggest, it is likely that the fringing field fromthe sensor captures the beads. A recent study by Ferreira et al. [39] shows that themagnetostatic fields created by the combined pinning and free layers in spin-valve sensorscontribute significantly to the magnetization of beads. In zero applied field Ferreira getsfield values comparable to the applied otherwise field (H = 1200 A m−1). Therefore themagnetostatic fields from planar Hall sensors should also be sufficient to attract beads,since the pinning layer is the same and the sensing layer is twice as thick as the spin-valvesensor’s free layer.

The fringing fields are not significant in magnetizing the beads detected by the planarHall sensor. The beads feeling this fringing field are situated at places, F1 and F2, wherethey are not detected by the planar Hall sensor. Beads inside the areas F1 and F2 givesmall contributions to the over-all average field, see Chapter 4. The effect of capturingbeads at the voltage leads should be avoided, nevertheless, due to loss of biologically activematerial. Maybe shielding by the SU-8 layer could be used for this purpose.

Finally, the field produced by the AC current through the sensor should be consideredin the theoretical analysis. When calculating the magnitude of this contribution withrespect to magnetizing the beads, one gets that it averages to zero. Thus only the appliedfield needs to be considered in the theoretical analysis, as is indeed the case of Eq. (4.33).

7.4 Conclusion

This chapter has provided a procedure for immobilizing antibodies on an SU-8 surface.The procedure is very simple and involves only PBS solutions, which are non-toxic andcheap.

Detection of monoclonal antibodies has been realized using the planar Hall sensors,Batch 2. Increasing antibody concentrations result in increasing sensor signals. The signalis positive because more beads are bound outside the sensor than on top of the sensor.

It can be concluded that the chosen SU-8 design is unfortunate since the signal dependson the difference in layer coverage between outside and on top of the sensor. However, theaim of the SU-8 design has been to avoid the positive signal from beads attached outsidethe sensor area, and the design is thus justified. That the antibodies bind equally well tothe residual SU-8 layer left after development has been realized too late in the process totake into account in the SU-8 design.

Another SU-8 design, where the positive contributions from beads outside the sensor


area are shielded by a thicker layer, would be optimal in this case. A closely packedmonolayer of beads would thus give a normalized signal of β ≈ −0.021, see Chapter 4.Another benefit would be the possibility of avoiding bead capture by the fringing fieldsfrom the voltage leads.

Chapter 8

Influenza

In Denmark an influenza epidemic strikes 5-7 times on 10 years, during which 10-30 % ofthe population is infected with the influenza virus and develops the disease. It is estimatedby Statens Serum Institut (SSI, www.ssi.dk) that influenza causes approximately 1000deaths per year. Especially elderly or weak people are subjected to increased death rates.

Influenza can be treated with antiviral drugs, but the treatment must be given to thepatient within 48 hours after infection to have an effect. The earlier the treatment is giventhe better result of alleviating the symptoms and reducing the duration of the illness. Forthis reason quick diagnostics is important.

Influenza exists in a large genetic variety, and to complicate things further many otherdiseases resemble influenza in their symptoms. Hence a diagnostic test is necessary inorder to prescribe adequate drugs. Besides enabling the prescription of the drug, a reliablediagnostic test will also enable clinical testing of the antiviral drug against influenza. I.e.whether the cure is effective or not, and how effective is it.

Furthermore, in case of a pandemic, a quick diagnostic test would be very helpful tomap the spread of the disease and to isolate infected individuals. A similar test wouldalso be a great help for diagnosing avian flu in deceased birds.

Existing diagnostic tests for influenza are PCR and a number of quick tests based onimmunochemistry. In contrast to the immunochemistry based tests, the PCR technique isreliable, but the sample must be treated carefully to avoid contamination. A contaminatedsample would give a false result. Furthermore, the PCR equipment is expensive andrequires skilled personnel for operation. Hence the transport of a sample to, e.g., SSI aswell as the reporting back after processing is time consuming. There is clearly need for aquick and reliable diagnostic test.

The existing quick influenza tests, based on immunochemistry, are unreliable due tothe following reasons:

1. Low concentration of vira in the sample.

2. Aggressive enzymes in the sample, which can break down the antibodies on the chip.

2 is often solved by diluting the sample, but this of course increases the problem of 1.Introduction of magnetic beads offers possible solutions to both problems.

111

112 CHAPTER 8. INFLUENZA

Magnetic beads and magnetic immunoassay

In a future application, the antigen can be captured from the sample with magnetic beadsand the rest of the components can subsequently be washed away. This approach will solvethe problems of 1 and 2 in one step. Afterwards, the presence of the antigen is measuredin a magnetic immunoassay, where instead of eluding the antigen into clean buffer, thecarrier magnetic bead is detected.

The long-term aim of the magnetic immunoassay is to combine the advantages ofimmunochemistry with those of magnetic detection:

1. Magnetic capturing. This step increases the antigen or virus concentration anddilutes unwanted molecules in the sample simultaneously.

2. Magnetic detection. The carrier magnetic bead is detected directly after virus cap-ture. The planar Hall sensors are very sensitive, it is theoretically possible to distin-guish a single bead. Hence very low virus or antigen concentrations can potentiallybe detected.

3. Unspecifically bound beads can potentially be removed with magnetic forces. Thiseliminates washing after sample incubation.

Chapter outline

This chapter demonstrates the magnetic immunoassay performed with two Batch 2 chips,each consisting of three planar Hall sensors. Development of the influenza-chip introducesthe medical treatment for influenza but the principle is general and can be expanded to in-clude numerous diseases as well as genetics, mutations, genetical diseases, etc. Monoclonalantibodies [38] are supplied by SSI.

8.1 Immunochemistry

The basic principle of immunochemistry is an indirect detection of an antigen by reactionwith a specific antibody. Specific antibody proteins, called capture antibodies, are immo-bilized on a predefined spot on the surface of a chip. The antigen and antibody fit like akey in a lock, and the sample molecules bind specifically to the capture antibodies. Af-terwards, labelled detector antibodies, which in most existing immunoassays are labelledwith enzymes or fluorescent molecules, bind specifically to the sample antigen. Opticaldetection of the fluorescence molecule reveals the presence of the detector antibody, andthe specific location reveals the identity of the antigen in case of a sample of multipleinvestigated molecules.

Antibodies are generally known from the immune system, where they play an importantrole in fighting intruding diseases. The Human IgG molecule, Fig. 8.1, is an example ofsuch an antibody. IgG’s immobilized on the surface have similar appearance and areusually drawn like a Y, where the antigen is attached to one of the ”arms”.

Fig. 8.2 shows a typical sandwich assay experiments. The capture antibody is immobi-lized on a solid support, antigen is added and binds to one of the binding sites. Then the

8.1. IMMUNOCHEMISTRY 113

Figure 8.1: Human IgG molecule. This image shows the functional version of the antibodyin human blood, where the binding sites capture foreign intruder molecules. The image isproduced by the US National Library of Medicine (http://www.nlm.nih.gov).

Figure 8.2: Immunochemistry. The capture antibody is immobilized on the surface. Thenthe antigen is attached to the immobilized antibody and reaction with the detector anti-body can take place. In this assay sketch the detector antibody is labelled with biotin,which binds tightly and specifically to streptavidin. Detection is thus made with fluores-cence labelled streptavidin or streptavidin coated magnetic beads.


detector antibody binds to the antigen. In this sketch the detector antibody is labelledwith biotin, which can be visualized by fluorescent streptavidin.

In a magnetic immunoassay, the fluorescence label is exchanged with a magnetic bead,and the presence of the antigen detected via the presence of the bead. This gives thepossibility of an electronic readout realized by the planar Hall sensor.

In the magnetic sandwich assay presented in this chapter, the detector antibodies areimmobilized directly on the surface of the magnetic beads. This way the extra biotin-streptavidin binding step is bypassed.

8.2 Fluorescent influenza immunoassay

This section presents the procedure and the results obtained for the fluorescent influenzasandwich immunoassay. First the tests on various blocking systems are summarized, basedon which BSA and milk is chosen. Then the assay procedure is described, and finally theresults are presented.

Blockers

A blocker is a solution of proteins that attach readily to the surface. These proteins areadsorbed at spaces where no antibodies are attached and occupy the space such that thedetector antibodies cannot attach to the surface. Thus the signal from the negative controlis decreased, and preferably identical to the background.

The different blocking systems that have been tested are:

1. BSA, cBSA = 5 mg ml−1 in PBS.

2. BSA + milk, cBSA = 5 mg ml−1 in PBS + 5 % fat free milk.

3. Rabbit serum, diluted 10 times.

4. Bovine serum, diluted 10 times.

For the fluorescent biotinylated antibody-streptavidin system (Chapter 7), BSA (1) isinsufficient to block the entire SU-8 surface. High background signals are observed. Theaddition of fat free milk powder (2) decreases the background. The lowest backgroundis obtained with diluted rabbit serum (3). Thus rabbit serum has been used as blockingsystem for a while. Bovine serum (4) has not been tested for this system.

When introducing the influenza sample to the detection system, the signal obtainedwith the rabbit serum (3) as blocker vanishes. Therefore all blockers are tested with theinfluenza system. The concluding results are: Rabbit and bovine serum (3 and 4) eraseall traces of influenza. BSA alone (1) is insufficient as blocking system, but BSA and milk(2) works well enough to get a significant signal from the influenza sample compared tothe negative control.

Thus BSA and milk (2) is chosen as blocking system for both the biotin-streptavidinsystem and the influenza assay.

8.2. FLUORESCENT INFLUENZA IMMUNOASSAY 115

8.2.1 Procedure

Fig. 8.2 is a schematic drawing of the sandwich immunoassay. As detection system Cy5labelled monoclonal antibodies are used. Biotinylated monoclonal antibodies are avoidedsince high unspecific binding is observed when using biotin-streptavidin as detection sys-tem. The influenza antibodies and antigens are provided by Statens Serum Institut (SSI).

1. Monoclonal antibodies are immobilized on the SU-8 surface. Concentrations usedare cmab = 0.1 mg ml−1, cmab = 0.5 mg ml−1, and cmab = 5 mg ml−1 in PBS.The positive control is cmab∗ = 0.1 mg ml−1 Cy5 labelled monoclonal antibodies,and the negative control is cmab = 5 mg ml−1 unmarked antibodies. A fat barrierseparates the controls from the samples. Procedure: 1 hour at T = 37 oC in a humidatmosphere.

2. The surface is blocked with BSA and milk. Blocking solution: cBSA = 5 mg ml−1 inPBS + 5 % fat free milk. Procedure: half an hour at room temperature.

3. Antigens are allowed to react with the antibodies overnight at T = 4 oC. Thenadditional antigens are placed on the sample spots and the slide is left 1 hour atT = 37 oC. The concentration is approximately c = 106 vira ml−1. Procedure: >10hours at T = 4 oC + 1 hour at T = 37 oC in a humid atmosphere.

4. The microscope slide is washed in BSA. cBSA = 5 mg ml−1 in PBS. Procedure: 1/2hour at room temperature.

5. Cy5 labelled monoclonal antibodies with a concentration of cmab∗ = 5 mg ml−1 inPBS are added on the total slide area including both sample spots and control spots.Procedure: 1 hour at T = 37 oC in a humid atmosphere.

6. Finally the slide is washed in PBS (procedure: 5 minutes in a beaker with manualstirring) and rinsed with DI water (removes the salt). After drying in a nitrogenflow, it is ready for scanning.

8.2.2 Results

The scanning results are shown in Fig. 8.3. The positive controls are clearly visible due tothe success of direct Cy5 labelling of antibodies. The negative controls are not visible, andthe intensity of their fluorescence signal equals the background. This proves that antibody-antibody binding does not occur. The signal from the sample spots is approximatelyfive times the signal from the negative control for the highest concentrations of captureantibodies.

Capture antibody concentration 0.1 mg ml−1 0.5 mg ml−1 5 mg ml−1

Signal sample/negative control 1.6 4.8 5.3Positive control/negative control > 100 > 100 > 100Negative control/background ≈ 1 ≈ 1 ≈ 1


Figure 8.3: Scanning results of the influenza immunoassay. The microscope slide is viewedfrom above, the slide is 2.5 cm wide (top to bottom in the image). The left column consistsof four clearly visible positive controls. To their immediate right are the negative controls,which are not visible. A fat barrier separates the controls from the samples. The threecolumns of sample spots are barely visible. The concentrations of capture antibodies forthe sample are from left to right: cmab = 5 mg ml−1, cmab = 0.5 mg ml−1, and cmab = 0.1mg ml−1 with the concentration from top to bottom unchanged.

8.2.3 Discussion

The signal compared with the negative control is on average 5 for the two highest captureantibody concentrations. This is a low signal. The virus concentration used is high(c = 106 vira ml−1). Ideally speaking, detection of a concentration of c = 103 vira ml−1

would be desirable for clinical trials.It will not be possible to detect c = 103 vira ml−1 using fluorescence scanning, since

the 1000 times lower signal would not be seen in the scanning results. The system isoptimized on SU-8, and signal enhancement is unlikely. Thus it can be concluded that thefluorescence detection system will probably fail in influenza diagnostic tests.

8.3 Magnetic influenza immunoassay

This section provides the procedure for the magnetic sandwich immunoassay including an-tibody adsorption onto beads and preparation of the planar Hall chip. Successful detectionis presented in Table 8.1, Fig. 8.4, and Fig. 8.5.

8.3.1 Procedure: antibody adsorption to beads

Antibodies can be adsorbed onto polystyrene readily and permanently [40]. The followingprocedure has been developed for adsorption of proteins onto polystyrene beads by BangsLaboratories, Inc. [40], and modified using the experiences from SSI.

1. Washing procedure. 50 µl of plain Nanomag-D beads (original concentration c = 25mg ml−1) are suspended in 500 µl of PBS. The beads are captured with a magnet

8.3. MAGNETIC INFLUENZA IMMUNOASSAY 117

and re-suspended in 500 µl fresh PBS three times. The bead concentration is now:c = 2.5 mg ml−1.

2. Mix monoclonal antibodies, cmab = 5 mg ml−1, and Nanomag-D beads cbeads = 2.5mg ml−1. Solution: 25 µl beads + 20 µl antibodies + 55 µl PBS.

3. Place the vial at T = 37 oC for 3 hours. Procedure: shake every 5 minutes duringthe first 20 minutes. Shake occasionally (every 20 or 30 minutes) for the remainingtime.

4. Wash three times in 100 µl PBS.

5. Use the beads immediately after preparation. Long term storage in PBS should beavoided due to eventual degradation of the antibodies.

8.3.2 Procedure: planar Hall chip

1. The signals from the three sensors, A, B, C, are measured. The drift is subtractedfrom the sample sensor by use of a reference sensor as described in Chapter 7, seeAppendix B for details on data acquisition and analysis.

2. Monoclonal antibodies are immobilized on the SU-8 on top of the planar Hall sensorsfor the sample and for the negative control. Sensor A is the sample, sensor B thereference, and sensor C is the negative control. The antibody concentration used isAssay 1: cmab = 5 mg ml−1 in PBS, and Assay 2: cmab = 0.5 mg ml−1 in PBS. A fatbarrier is placed on top of sensor B to avoid contamination of the reference sensor.Procedure: 1 hour at T = 37 oC in a humid atmosphere.

3. The remaining surface is blocked with BSA and milk. Blocking solution: cBSA = 5mg ml−1 in PBS + 5 % fat free milk. Procedure: half an hour at room temperature.

4. Antigens are placed on sensor A and allowed to react with the surface antibodies.cvirus ≈ 106 ml−1. Procedure: 1 hour at T = 37 oC in a humid atmosphere.

5. The chip is washed in BSA. Blocking solution: cBSA = 5 mg ml−1 in PBS. Procedure:half an hour at room temperature.

6. Magnetic beads with monoclonal antibodies are added to sensor A and sensor C.Procedure: 1 hour at T = 37 oC.

7. Finally the chip is washed 5 minutes in PBS and left overnight in order to be com-pletely dry before proceeding with the electrical measurements.

8. The signals from the three sensors are measured, and the reference signal is sub-tracted from the sample and the negative control. Comparison with the previoussignal gives the signal from the beads, ∆βbeads = ∆βafter −∆βbefore.


Assay ccab (mg ml−1) ∆βsample (×10−3) ∆βnegative control (×10−3) S/NC

1 5 4.7(0.5) 3.0(0.4) 1.62 0.5 5.9(0.3) 2.4(0.9) 2.5

Table 8.1: Magnetic immunoassay results. ccab is the concentration of capture antibodies,∆β the measured signal from the beads (normalized to the applied field), and the numbersin parentheses are the experimental uncertainties. S/NC is calculated as the signal fromthe assay divided by the signal from the negative control.

8.3.3 Results

Two experiments are performed with the influenza assay, and the results are presented inTable 8.1. Assay 1 is constructed with a capture antibody concentration of cmab = 5 mgml−1, and Assay 2 with a capture antibody concentration of cmab = 0.5 mg ml−1.

Fig. 8.4 and Fig. 8.5 compare the results from the influenza assay with those from thereference assay. The influenza assay is plotted on top of the reference assay (Fig. 7.10,Chapter 7) to obtain a rough estimate of the magnitude of unspecific binding. Additionally,to verify that the signal from the sample does not exceed the saturation value from thereference curve.

In both figures, the top graph shows the electrical results with the sample in red andthe negative control in blue, the grey points are the reference measurements of signal asa function of antibody concentration, Fig. 7.10. The bottom row shows two pictures ofthe sensors taken just after the biochemical process has been completed, the left is thenegative control sensor and the right is the sample sensor.

In both experiments the signal of the sample can be clearly distinguished from thesignal of the negative control. In Assay 1 the difference between the sample and thenegative control is visible in the pictures of the sensors with beads, whereas the picturesof Assay 2 show less clear differences.

The S/NC of the assay, measured as the signal from the sample divided by the signalfrom the negative control, is S/NC = 1.6 (Assay 1) and S/NC = 2.5 (Assay 2). This ison average S/NC = 2.

8.4 Discussion

The influenza magnetic sandwich immunoassay presented in Table 8.1, Fig. 8.4, andFig. 8.5 captures the presence of the virus. In both cases, the electrical measurementsare clearly able to distinguish between the sample and the negative control.

Looking at the pictures of the sensor just after the biochemistry has been performed itis not clear why the largest difference between sample and negative control is observed inAssay 2. The higher signal from Assay 2 can be attributed to a combination of more beadsplaced at positively contributing sites and less beads placed at negatively contributingsites. The analysis of bead signal in Chapter 4 supports this postulate. The analysisshows that a bead’s position with respect to the sensor edge can have a strong influence

8.4. DISCUSSION 119

-10 0 10 20 30 40 50 60 70 80 90 100 110

0.000

0.001

0.002

0.003

0.004

0.005

0.006

0.007Influenza Assay 1

Negative control

Influenza sample

afte

r-be

fore


Figure 8.4: Results of influenza Assay 1 with cmab = 5 mg ml−1 capture antibody con-centration. Top graph shows the electrical measurements with the sample (red) clearlydistinguishable from the negative control (blue). The c-values can be estimated from thereference measurements of signal versus antibody concentration (grey) to verify that sam-ple and negative control fall within the boundaries of the reference. The pictures of thesensor in the bottom row are taken just after the biochemical experiments. Left: thenegative control. Right: the sample.


-10 0 10 20 30 40 50 60 70 80 90 100 110

0.000

0.001

0.002

0.003

0.004

0.005

0.006

0.007Influenza Assay 2

Negative control

Influenza sample

afte

r-be

fore


Figure 8.5: Results of influenza Assay 2 with cmab = 0.5 mg ml−1 capture antibody con-centration. Top graph shows the electrical measurements with the negative control (blue)clearly below the sample (red). The c-values can be estimated from the reference mea-surements of signal versus antibody concentration (grey) to see that sample and negativecontrol fall within the boundaries of the reference. The pictures of the sensor in the bottomrow are taken just after the biochemical experiments. Left: the negative control. Right:the sample.

8.5. CONCLUSION 121

on the over-all signal. In this way Assay 2 contrasts Assay 1, where the picture clearlydistinguishes between the sample and the negative control.

The signal to negative control of the two magnetic influenza assays are S/NC = 1.6(Assay 1) and S/NC = 2.5 (Assay 2), on average S/NC = 2. This is of course notvery impressive compared to approximately S/NC = 5 of the fluorescent sandwich assay(Fig. 8.3), but as a first result it is promising.

The signal from the negative control is ∆β = 0.002-0.003 (normalized to the appliedfield), corresponding to an antibody concentration of cmab = 20-30 µg ml−1. The flu-orescent signal from cmab = 30 µg ml−1 antibodies reaches approximately 75 % of thesaturation signal obtained for biotinylated monoclonal antibodies in the Cy3-streptavidindetection system (Fig. 7.2). This is very high for a sandwich assay and relates to theproblem of unspecific binding and/or bead adsorption to proteins or field gradients on thechip. Solving this particular issue of the magnetic immunoassay would improve the resultsconsiderably.

The possibility of an electrical signal-to-noise enhancement of 50-100 has been proposedin Chapter 6. The estimate is based on improving several features of the experimentalsetup. With these improvements the signal-to-noise from a single 250 nm Nanomag-Dbead can amount to S/N ≈ 2-3 (compared to Table 4.3) offering great sensitivities onimmunoassays. However, the possibility of distinguishing the presence of a single bead canonly be fully exploited when the unspecific binding from the negative control is solved.

8.5 Conclusion

This chapter has presented a detection procedure for influenza based on immunochem-istry. The detection of proteins with antibodies immobilized on an SU-8 surface has notpreviously been explored to the author’s present knowledge. The procedure of influenzadetection has been optimized with respect to surface blocking system and the fluorescentdetection system, using direct labelling of detector antibodies.

The results of the influenza assay have been obtained first with fluorescently labelledmonoclonal antibodies, yielding S/NC = 5, then with the Batch 2 planar Hall sensors,S/NC = 2, where S/NC is the signal from the sample divided by the signal from thenegative control.

The successful detection of influenza by the planar Hall sensors is promising. The highsignal from the negative control gives a low assay sensitivity. Decreasing the unspecificbinding will thus improve the assay result considerably. Also the capturing of beads atundesired places could possibly be avoided by SU-8 design considerations.

Chapter 9

Conclusion and outlook

9.1 Conclusion

Planar Hall sensors are promising candidates for a prototype diagnostic chip as they canbe used for point-of-care diagnostics of influenza infection. Their inherent signal-to-noisecharacteristic outperforms competitive sensing principles, at least for DC detection mea-surements. Planar Hall sensors have proven capable of detecting relatively low antibodyconcentrations especially considering the SU-8 capping design, which result in competingpositive and negative signals. Additionally, actual influenza sandwich immunoassays havebeen detected with the studied planar Hall sensors.

Design, fabrication and characterization of planar Hall sensors of different sensingmaterials have been investigated theoretically and experimentally. In the end, the optimalmaterial for planar Hall sensors is chosen for the immunoassay experiments. This is thetop pinned permalloy stack. However, the optimal design for the SU-8 capping layer isnot put to the experimental test. This design has been realized too late in the process toimplement in the fabrication.

For the electrical characterization of sensors, a setup has been built that eliminatesmuch of the electrical noise observed in the first batch of sensors [4, 5]. The noise isreduced almost by a factor of 100 in the new setup. The experimental noise of the setupis dominated by the internal noise of the lock-in amplifier, and it would be troublesometo extract the experimental noise from the sensor itself. However, the introduction of alow-noise pre-amplifier between the sensor and the lock-in amplifier would help to utilizethe full scale of the planar Hall sensor. An estimate of the signal-to-noise enhancementfollowing this course is approximately 5. Combined with a factor of 10-20 from increasingthe sensor current (requires rebuilding of the experimental setup), gaining a factor of 50-100 is possible. An S/N enhancement of 60 is sufficient for single 250 nm Nanomag-Dbeads to be easily distinguished with the actual sensor batch.

The theoretical prediction of the sensor resisitvity, AMR, sensitivity, and temperaturedrift accounts reasonably for the experimentally observed behavior (presented in Table6.4). The same applies to the theoretical versus experimental signal from a monolayer ofbeads. The theoretical value for the studied sensors is β ≈ 0.024(ξoutside−ξsensor), where ξ

123

124 CHAPTER 9. CONCLUSION AND OUTLOOK

is the bead coverage, and the experimental maximum value is ∆βbeads = 0.011, implyingthe order of magnitude is correct.

The observed signals of magnetic immunoassays in Fig. 7.8, Fig. 7.10, and Table 8.1,range between ∆βbeads = 0 and ∆βbeads = 0.011. An increase in antibody concentrationresults in an increase in sensor signal. This is anticipated since more beads are expectedto bind to the biotinylated antibodies, when their surface concentration is increased.

An experimental procedure for immobilizing antibodies on SU-8 has been establishedduring this Ph.D. study. Various blockers has been investigated and the most promisingresults were obtained with the well-known combination of BSA and milk. Additionally, thefull protocol of an influenza immunoassay has been produced and optimized with respectto time consumption. The author acknowledges that more efforts could have been put intothis particular part of the study if time had allowed. Both procedures have been testedon the planar Hall sensors. The immobilization procedure of biotinylated monoclonalantibodies is studied with the planar Hall sensor as a function of antibody concentration(Fig. 7.8 and Fig. 7.10). Additionally, the influenza immunoassay is detected with theplanar Hall sensors with an average S/NC = 2, measured as the signal from the sampledivided by the signal from the negative control.

The obtained results are promising for implementing the planar Hall sensor in magneticimmunoassaying. Several improvements can be introduced, where particularly the SU-8design should lead to improvements. Furthermore, when the large signal from the negativecontrol is diminished, the resulting assay S/NC = 2 would be enhanced.

9.2 Outlook

This outlook provides the author’s assessment of promising ways to pursue the workfurther. There is scope for further improvement in both sensor design and bead design,which is a prerequisite for the launch of a prototype device. The two main issues aredescribed below.

9.2.1 Sensor technology

The size of the sensor can be tuned to match the application. For immunoassay detection,smaller sensors seem preferable to larger ones. In this case, sensitive areas of 20µm×20µmhave larger signals from the respective antibody concentrations compared with areas of1mm × 1mm (data not shown). One should also investigate the theoretical signal fromsensitive areas differing from square geometry. It might be possible to cut some positivecontributions and optimize the sensor design with respect to the bead signal.

Another issue to look into is the design of the SU-8 layer covering the sensor. The designused in the work presented here represents a first step, which can be improved considerably.The focus of the current design was to ensure immobilization of antibodies only above thesensor area. However, the antibodies bind just as well to the SU-8 monolayer still presentafter development, and as a result the beads cover everything. Additionally, the beads getstuck at the edges of the SU-8 pad. They also have a tendency to stick to the edges of thepermalloy layers at locations with high fringing fields.

9.2. OUTLOOK 125

Instead of only covering the sensitive parts of the sensor, a thick layer of SU-8 can beused to block the signal from the beads everywhere else than on top of the sensor. Leavingsmall grooves or wells on top of the sensor will allow easier placement of antibodies as wellas shield the positive contributions from the beads. Another benefit will be to shield thelargest part of the fringing field from the permalloy edges and thus limit the tendency ofbeads to stick in undesired locations.

9.2.2 Bead design

For immunoassays and DNA array detection it is essential to investigate further the aboveissues on sensor technology. However, a large number of researchers are involved in de-veloping better and more sensitive systems for detection while an area, which is presentlyless in focus, is the design of beads for diagnostic purposes. The beads detected in thisthesis are designed for separation, not for diagnostics, and some of the troubles encoun-tered during this study could be directly related to this fact. Non-specific binding is oneof the greatest issues. If quantitative measurements are desired, unspecific binding of thebeads must be dealt with and preferably fully avoided.

One thing, which is feasible, is to go down in bead size. In this approach gravity willhave less impact on the bead’s settlement on the surface. This ought to be combined withhigher magnetic moment, if bead sizes become very small. Alternatively, larger bead sizescould be chosen and in that way larger viscous forces are induced on the beads duringwashing.

It is also advisable to study the impact different buffers have on unspecific binding.For instance, adding a small amount of Triton X-100 to the bead solution might decreaseunspecific binding. Also adjusting the pH of the buffer could have an effect if unspecificbinding is related to electrostatic forces. Quantitative experiments have to be performedto fully understand the effects.

Another issue to address is how to avoid agglomeration when fishing antigens out ofa solution. One virus with antigens on the surface can bind to more than one antibodycovered bead and thus when concentrating the reagents by magnetic forces, long chainsof bead-virus-bead-virus can easily form. This should be taken into account in the beaddesign or the preparation of the sample.

126 CHAPTER 9. CONCLUSION AND OUTLOOK

Appendix A

Analytical expressions

A.1 Analytical solution to the homogeneously magnetizedsphere

Consider a sphere of linear permeability placed in an external field H0x. The field variationin the xy-plane at a given distance, z, is of interest to the planar Hall sensor geometry. Inthe case of no free currents Maxwell’s equations can be reduced to Laplace’s equation fora magnetic potential, U(r, θ), defined as B = −∇U(r, θ).

∇2U(r, θ) = 0 (A.1)

In spherical coordinates with no φ-dependence Eq. (A.1) becomes

∇2U(r, θ) =1r2

∂

∂r

(r2 ∂U(r, θ)

∂r

)+

1r2 sin θ

∂

∂θ

(sin θ

∂U(r, θ)∂θ

)= 0 (A.2)

The high symmetry of the solid sphere and the external field implies that the potentialcan be separated into a term depending only on r, R(r), and a term depending only on θ,Θ(θ). Inserting U(r, θ) = R(r)Θ(θ) into Eq. (A.2) and multiplying by r2 yields

∇2U(r, θ) =∂

∂r

(r2 ∂R(r)

∂r

)Θ(θ) +

1sin θ

∂

∂θ

(sin θ

∂Θ(θ)∂θ

)R(r) = 0 (A.3)

Dividing by U(r, θ) and rearranging Eq. (A.3) one gets the following differential equation

1R(r)

∂

∂r

(r2 ∂R(r)

∂r

)= − 1

Θ(θ) sin θ

∂

∂θ

(sin θ

∂Θ(θ)∂θ

)(A.4)

It is obvious from Eq. (A.4) that the LHS depends only on r and the RHS depends onlyon θ. This can only be fulfilled if both sides equals the same constant which without lossof generality can be called l(l + 1). First the LHS is solved, the differential equation is

d

dr

(r2 dR(r)

dr

)= l(1 + l)R(r) (A.5)

127

128 APPENDIX A. ANALYTICAL EXPRESSIONS

Eq. (A.5) has the general solution

R(r) = Crl +D

rl+1(A.6)

which can be easily verified by insertion into the differential equation. Second, the RHSof Eq. (A.4) is considered

∂

∂θ

(sin θ

∂Θ(θ)∂θ

)= −l(l + 1) sin θΘ(θ) (A.7)

The general solution to Eq. (A.7) is the Legendre polynomials

Θ(θ) = Pl(cos θ) (A.8)

Pl(x) =1

2ll!

(d

dx

)l

(x2 − 1)l

dPl

dx=

12ll!

(d

dx

)l+1

(x2 − 1)l

The general solution to Eq. (A.4) is the product of Eq. (A.6) and Eq. (A.8)

U(r, θ) =∞∑

l=0

(Clr

l +Dl

rl+1

)Pl(cos θ) (A.9)

Inside the sphere (r ≤ R, where R is the radius of the sphere) the potential should befinite requiring D(r≤R) = 0, and outside the sphere (r ≥ R) the potential should vanish atinfinity requiring C(r≥R) = 0. Summarizing these requirements,

U(r, θ) =

∑∞l=0 Clr

lPl(cos θ) r ≤ R,∑∞l=0

Dl

Rl+1 Pl(cos θ) r ≥ R.(A.10)

The potential should be continuous at r = R

∞∑

l=0

ClRlPl(cos θ) =

∞∑

l=0

Dl

Rl+1Pl(cos θ) (A.11)

for all l. This gives one constraint on Cl and Dl (Dl = ClR2l+1). The gradient, on

the other hand, is discontinuous at the surface of the sphere, where the discontinuity isproportional to the normal part of the magnetization, µ0M,

(∇Uabove −∇Ubelow) · n = −µ0M · n (A.12)

∂Uabove

∂r|r=R − ∂Ubelow

∂r|r=R = −µ0M cos θ

A.1. 129

∂U

∂r=

∑∞l=0 Clr

l−1Pl(cos θ) r ≤ R,∑∞l=0−(l + 1)ClR

2l+1

rl+2 Pl(cos θ) r ≥ R.

∂Uabove

∂r|r=R − ∂Ubelow

∂r|r=R =

∞∑

l=0

−ClPl(cos θ)Rl−1(2l + 1)

∞∑

l=0

ClPl(cos θ)Rl−1(2l + 1) = µ0M cos θ (A.13)

Because the Legendre polynomials are orthogonal, using Fourier’s trick will simplify theproblem considerably. The orthogonality relations of the Legendre polynomials are

∫ 1

−1Pl(x)Pm(x)dx =

0 l 6= m,

22l+1 l = m.

Multiplication on both sides of Eq. (A.13) with Pm(cos θ) and integration over all spacespanned by the functions yields

∫ 1

0

∞∑

l=0

ClPl(cos θ)Rl−1(2l + 1)Pm(cos θ)d(cos θ) =∫ 1

0µ0M cos θPm(cos θ)d(cos θ)

Cm =12R1−mµ0M

∫ 1

−1cos θPm(cos θ)d(cos θ)

= −12R1−mµ0M

∫ π

0cos θPm(cos θ) sin θdθ

Since P1 = cos θ

Cm =

µ0M

3 m = 1,

0 m 6= 1.(A.14)

Finally, Eq. (A.14) can be inserted into the solution Eq. (A.10) to obtain the final solution

U(r, θ) =

µ0M

3 r cos θ r ≤ R,µ0M

3R3

r2 cos θ r ≥ R.(A.15)

The magnetic flux density is obtained by taking the gradient of the potential Eq. (A.15).

B = −∇U

Outside the sphere the well known result of the magnetic dipole field [32] is obtained

B =µ0M

3R3

r3(2 cos θr + sin θθ) (A.16)

And the magnetic field strength outside the sphere is

H =M

3R3

r3(2 cos θr + sin θθ) (A.17)

130 APPENDIX A. ANALYTICAL EXPRESSIONS

Appendix B

Data acquisition and analysis

B.1 Data acquisition

Temperature IAC f τ (lock-in) Slope (lock-in) Phase (lock-in)30 oC 1 mA 330 Hz 100 ms 12 dB oct−1 0 deg

Table B.1: Experimental settings of the lock-in amplifier during antibody detection ex-periments. IAC is the current (maximum amplitude), f is the frequency, and τ is the timeconstant.

1. A chip with three sensors, A, B, and C, is placed in an electrically shielded box.The temperature of the whole box is controlled as described in Chapter 7 at atemperature of T = 30 oC. Thermal equilibrium of the box happens fast but thetransfer of thermal equilibrium to the inside of the box must be assumed slower.Therefore the temperature of the box is always kept at T = 30 oC and the change ofchips carried out as quickly as possible in order to minimize heat exchange with thesurroundings. Immediately after mounting the chip in the box, the lid is closed andthe current applied to all sensors in a series connection. The current is IAC = 1 mA(maximum amplitude) with a frequency of f = 330 Hz. The current is applied toall sensors in series and does not change throughout the entire measurement. Withcurrent applied, a waiting period of 1 hour is conducted for everything to stabilize.This includes equilibrium temperature and sensor offset.

2. The rest of an experiment is controlled via LabView.

3. First the external magnetic field is set to B = +0.5 mT. Then sensor A is chosen bythe switch box and LabView waits 10 s before the experiment continues. The timeconstant is set to 100 ms, the slope is 12 dB oct−1. The sensitivity of the lock-inmatches this particular chip and sensor and is set prior to the measurement. Typicalsensitivities are 50-100 µV.

131

132 APPENDIX B. DATA ACQUISITION AND ANALYSIS

4. The signal from sensor A is measured 100 times with the time constant 100 ms andthe slope 12 dB oct−1. Afterwards, the switch box switches to sensor B, and aftera waiting time of 1 s the signal from sensor B is measured 100 times. Finally, theswitch box switches to sensor C and after waiting 1 s the signal from sensor C ismeasured 100 times.

5. Then sensor A is switched back on and the sequence of 3 and 4 is repeated 50 timesin total.

6. The magnetic field is changed from B = +0.5 mT to B = −0.5 mT. Sensor A isthen chosen. Again LabView waits 10 s before proceeding. The measurements arerepeated exactly as they were for the positive field, see 3-5.

The whole experiment takes 1 hour to complete, half an hour at each field value. Afterthis half hour the signal is still stabilizing towards the specific field value. This approach-ing an equilibrium value is related to the change in magnetic field. Nothing else changes.Additionally, a difference between plus and minus the field value is observed which corre-sponds to the smaller and larger jumps in field. See Fig. B.1, left graph. When changingto B = −0.5 mT, the field is changed by a total amount of ∆B = 1 mT (from B = +0.5mT) whereas at B = +0.5 mT the change is only ∆B = 0.5 mT (from B = 0 mT). Hencethe difference between the first and last data points is larger when changing the field fromB = +0.5 mT to B = −0.5 mT than it is when changing from B = 0 mT to B = +0.5mT.

B.2 Data analysis

The signal from sensor A in B = +0.5 mT, VA(+0.5 mT), and the sensitivity, SA, havearbitrary sign from chip to chip. All sensitivities on the same chip, SA, SB, SC, havealways the same sign but the offset might be large enough to change the sign of the actualsignal in B = +0.5 mT internally on a chip. Since the phase can vary from one sensor toanother, the signal reported is the R value (the norm of the vector). This value is alwayspositive even when X changes sign. Therefore the sign of X is observed along with R. Iftwo R values have corresponding X values with different signs, the sum is taken (R1+R2)when the difference is desired (X1-X2). Remember, that the behavior of X is of interestbut R is used in order to avoid correcting for the phase.

1. The signal from one sensor is normalized to the sensitivity of that particular sensorand the current through the sensor. This yields a value related to the magnetic fluxdensity felt by sensor A, BA:

BA =VA

SA · IAC (mA)(B.1)

with the unit of mT. In Eq. (B.1) VA is the measured voltage of sensor A, SA is thesensitivity of sensor A, and IAC is the applied current (maximum amplitude). Thescaled version of the data representation is shown in Fig. B.1, right graph.

B.2. DATA ANALYSIS 133

2. When dividing by the applied flux density, B (measured in mT), a dimensionlessnumber is obtained, β, which still reflects the magnetic field experienced by thesensor. Exactly the same arguments hold for the other sensors and we are left withthree dimensionless values: βA, βB, and βC.

βA =VA

SA · IAC (mA) ·B (mT)(B.2)

3. The off-set VA(0 mT) stabilizes after approximately one hour with current appliedto the sensor. Therefore a waiting time of one hour after mounting the sensorand applying the current is necessary to obtain correct results. The off-set canchange considerably with demounting and remounting the sample. It also changeswith temperature. Therefore the temperature is kept constant allowing for off-setcorrection by taking the mean value at negative and positive fields.

∆βA =12βA(+0.5 mT) +

12βA(−0.5 mT) = (B.3)

VA(+0.5 mT)SA · 1 mA · 1 mT

+VA(−0.5 mT)

SA · 1 mA · (−1 mT)≈ 1

yielding a number close to one, see Fig. B.2, left graph.

4. Sensor B is always the reference. This sensor is never changed during the biochemicalreactions and will thus always have a β-value close to one. Sensor A and sensor Cmeasure the applied field plus the field from the beads on top of the respectivesensor. Since the field from the beads is negative, their β-values are reduced due tothe presence of beads.

5. Subtracting the reference yields

∆βA,C −∆βB

= 0 without beads,< 0 with beads.

(B.4)

In Eq. (B.4) ”0” is to be understood as baseline since asymmetries can be connectedwith the normalizing procedure. The resulting data representation for sensor Awithout beads is shown in Fig. B.2, right graph.

6. With a very tiny amount of beads on top of sensor C, the ∆β-value drops as expected,see Fig. B.3, left graph.

Investigation of reference stability

For a large number of measurements the reference sensor gives a stable output. During themeasurements shown in Fig. B.3, right graph, the sensor socket has broken and is gluedback to the box. This repair has no measurable effect on the reference ∆β-value.

Also the introduction of beads to the chip does not alter the signal from the reference.Beads contaminating the reference has been a problem, but this is avoided simply byplacing a fat barrier on top of the reference sensor. The fat barrier repels water and thusthe bead containing solution does not come into contact with the reference.

134 APPENDIX B. DATA ACQUISITION AND ANALYSIS

0 20 40 60 80 10022.85

22.90

22.95

23.00

23.05

23.10

23.15

23.20

23.25

23.30-0.5 mT+0.5 mT

Chip 202 Sensor A

VA

Demount

Sig

nal (

µV)

Measurement0 20 40 60 80 100

0.512

0.514

0.516

0.518

0.520

0.522-0.5 mT+0.5 mT

Chip 202 Sensor A

VA

Demount

BA (m

T)

Measurement

Figure B.1: Relaxation of the sensor signal in plus and minus applied field. Left: Measureddata (µV). Right: Scaled with the sensitivity and current through the sensor (mT).

0 10 20 30 40 50

1.0291.0301.0311.0321.0331.0341.0351.0361.0371.0381.0391.0401.0411.0421.0431.044

Chip 202 Sensor A. Off-set correction

Sensor A Demount

A(+

0.5m

T)-

A(-

0.5m

T)

Measurement0 10 20 30 40 50

0.0020

0.0025

0.0030

0.0035

0.0040Chip 202 Sensor A. Reference subtraction

Sensor A Demount

A-

B

Measurement

Figure B.2: Normalized off-set corrected sensor signal; the data representation is thedimensionless number β. Left: Off-set corrected values. Right: Off-set corrected andreference subtracted.

0 10 20 30 40 50

-0.003

-0.002

-0.001

0.000

0.001

0.002 Chip 202 Sensor C. Off-set and reference corrected

Sensor C Demount Beads Beads demountA

-B

Measurement0 10 20 30 40 50

1.010

1.012

1.014

1.016

1.018

1.020

1.022

1.024

1.026

Chip 203 Sensor B (reference)

Signal Demount Repair Demount after rep Reference Beads, demount

B

Measurement

Figure B.3: Bead detection experiment. Left: The ∆β-value drops as expected when beadsare present on top of the sensor. Right: Reference stability. The ∆β-value is unchangedin effect of demounting the sensor and repairing the socket.

Appendix C

Mathematica code

135

136 APPENDIX C. MATHEMATICA CODE

BEAD INTEGRALS.txt

CHAPTER 4FIG 4.7

Small sensorX=-2+4*i/40Y=-2+4*j/40In[5]:=F=(2*(x-X)^2-(y-Y)^2-1)/((x-X)^2+(y-Y)^2+1)^(5/2)/4In[6]:=For[i=1,i<41,i++,For[j=1,j<41,j++,m[i,j]=NIntegrate[F,x,-1,1,y,-1,1]]]In[7]:=Q=Array[m,40,40]In[8]:=ListPlot3D[Q]Export["C:\xy_plotax.eps",%]In[9]:=ListContourPlot[Q]Export["C:\xy_contour.eps",%]

FIG 4.8 Along x

X=40*i/1000Y=0F=(2*(x+X)^2-(y+Y)^2-1)/((x+X)^2+(y+Y)^2+1)^(5/2)For[i=1,i<1000,i++,m[i]=NIntegrate[F,x,-20,20,y,-20,20]]Q=Array[m,999]Export["C:\alongx_y1.dat",Q]ListPlot[Q]Export["C:\Data\Math\x_plot_alongx.eps",%]

Along yX=0Y=40*i/1000F=(2*(x+X)^2-(y+Y)^2-1)/((x+X)^2+(y+Y)^2+1)^(5/2)For[i=1,i<1000,i++,m[i]=NIntegrate[F,x,-20,20,y,-20,20]]Q=Array[m,999]Export["C:\alongx_y1.dat",Q]ListPlot[Q]Export["C:\Data\Math\x_plot_alongx.eps",%]

TABLE 4.5 C

X=20*i/100-0.2Y=20*j/10-2L=0F=(2*(x+X)^2-(y+Y)^2-1)/((x+X)^2+(y+Y)^2+1)^(5/2)For[i=0,i<83,i++,For[j=0,j<15,j++,K[0,j]=0,K[i,0]=0]]For[i=1,i<83,i++, For[j=1,j<15,j++,m[i,j]=NIntegrate[F,x,-20,20,y,-20,20]]]Q=Array[m,81,13]For[i=1,i<82,i++, For[j=1,j<14,j++,K[i,j]=0.25*(m[i,j]+m[(i+1),j]+m[i,(j+1)]+m[(i+1),(j+1)])]]For[i=1,i<82,i++, For[j=1,j<14,j++,L=L+K[i,j]]]L-182.283ListPlot3D[Q]Export["C:\Data\Math\xy_plot_C.eps",%]ListContourPlot[Q]Export["C:\Data\Math\xy_contour_C.eps",%]ListDensityPlot[Q]Export["C:\Data\Math\xy_density_C.eps",%]

Ei

Page 1

137

BEAD INTEGRALS.txtX=20*i/100+16-0.2Y=20*j/10-2L=0F=(2*(x+X)^2-(y+Y)^2-1)/((x+X)^2+(y+Y)^2+1)^(5/2)For[i=0,i<23,i++,For[j=0,j<15,j++,K[0,j]=0,K[i,0]=0]]For[i=1,i<23,i++, For[j=1,j<15,j++,m[i,j]=NIntegrate[F,x,-20,20,y,-20,20]]]Q=Array[m,21,13]For[i=1,i<22,i++, For[j=1,j<14,j++,K[i,j]=0.25*(m[i,j]+m[(i+1),j]+m[i,(j+1)]+m[(i+1),(j+1)])]]For[i=1,i<22,i++, For[j=1,j<14,j++,L=L+K[i,j]]]L-143.844

EoX=20*i/100+20-0.2Y=20*j/10-2L=0F=(2*(x+X)^2-(y+Y)^2-1)/((x+X)^2+(y+Y)^2+1)^(5/2)For[i=0,i<20,i++,For[j=0,j<15,j++,K[0,j]=0,K[i,0]=0]]For[i=1,i<20,i++, For[j=1,j<15,j++,m[i,j]=NIntegrate[F,x,-20,20,y,-20,20]]]Q=Array[m,18,13]For[i=1,i<19,i++, For[j=1,j<14,j++,K[i,j]=0.25*(m[i,j]+m[(i+1),j]+m[i,(j+1)]+m[(i+1),(j+1)])]]For[i=1,i<19,i++, For[j=1,j<14,j++,L=L+K[i,j]]]L126.033

F1

X=20*i/100-0.2Y=20*j/10+22L=0F=(2*(x+X)^2-(y+Y)^2-1)/((x+X)^2+(y+Y)^2+1)^(5/2)For[i=0,i<120,i++,For[j=0,j<11,j++,K[0,j]=0,K[i,0]=0]]For[i=1,i<120,i++, For[j=1,j<11,j++,m[i,j]=NIntegrate[F,x,-20,20,y,-20,20]]]Q=Array[m,118,9]For[i=1,i<119,i++, For[j=1,j<10,j++,K[i,j]=0.25*(m[i,j]+m[(i+1),j]+m[i,(j+1)]+m[(i+1),(j+1)])]]For[i=1,i<119,i++, For[j=1,j<10,j++,L=L+K[i,j]]]L-32.9518

F2X=20*i/100+23.2Y=20*j/10+22L=0F=(2*(x+X)^2-(y+Y)^2-1)/((x+X)^2+(y+Y)^2+1)^(5/2)For[i=0,i<86,i++,For[j=0,j<11,j++,K[0,j]=0,K[i,0]=0]]For[i=1,i<86,i++, For[j=1,j<11,j++,m[i,j]=NIntegrate[F,x,-20,20,y,-20,20]]]Q=Array[m,84,9]For[i=1,i<85,i++, For[j=1,j<10,j++,K[i,j]=0.25*(m[i,j]+m[(i+1),j]+m[i,(j+1)]+m[(i+1),(j+1)])]]For[i=1,i<85,i++, For[j=1,j<10,j++,L=L+K[i,j]]]L9.81117

Page 2


BEAD INTEGRALS.txt F3

X=20*i/100+23.2Y=20*j/10-2L=0F=(2*(x+X)^2-(y+Y)^2-1)/((x+X)^2+(y+Y)^2+1)^(5/2)For[i=0,i<86,i++,For[j=0,j<15,j++,K[0,j]=0,K[i,0]=0]]For[i=1,i<86,i++, For[j=1,j<15,j++,m[i,j]=NIntegrate[F,x,-20,20,y,-20,20]]]Q=Array[m,84,13]For[i=1,i<85,i++, For[j=1,j<14,j++,K[i,j]=0.25*(m[i,j]+m[(i+1),j]+m[i,(j+1)]+m[(i+1),(j+1)])]]For[i=1,i<85,i++, For[j=1,j<14,j++,L=L+K[i,j]]]L140.949

FIG 4.11+4.12Z0=0.125

CX=0.1*i-0.1Y=j-1Z=0.125L=0F=(2*(x-X)^2-(y-Y)^2-Z^2)/((x-X)^2+(y-Y)^2+Z^2)^(5/2)For[i=0,i<83,i++,For[j=0,j<15,j++,K[0,j]=0,K[i,0]=0]]For[i=1,i<83,i++, For[j=1,j<15,j++,m[i,j]=NIntegrate[F,x,-10,10,y,-10,10]]]Q=Array[m,81,13]For[i=1,i<82,i++, For[j=1,j<14,j++,K[i,j]=0.25*(m[i,j]+m[(i+1),j]+m[i,(j+1)]+m[(i+1),(j+1)])]]For[i=1,i<82,i++, For[j=1,j<14,j++,L=L+K[i,j]]]L-370.644

EiX=0.1*i-0.1+8Y=j-1Z=0.125L=0F=(2*(x-X)^2-(y-Y)^2-Z^2)/((x-X)^2+(y-Y)^2+Z^2)^(5/2)For[i=0,i<23,i++,For[j=0,j<15,j++,K[0,j]=0,K[i,0]=0]]For[i=1,i<23,i++, For[j=1,j<15,j++,m[i,j]=NIntegrate[F,x,-10,10,y,-10,10]]]Q=Array[m,21,13]For[i=1,i<22,i++, For[j=1,j<14,j++,K[i,j]=0.25*(m[i,j]+m[(i+1),j]+m[i,(j+1)]+m[(i+1),(j+1)])]]For[i=1,i<22,i++, For[j=1,j<14,j++,L=L+K[i,j]]]L-513.12

EoX=0.1*i+10-0.1Y=j-1Z=0.125L=0F=(2*(x-X)^2-(y-Y)^2-Z^2)/((x-X)^2+(y-Y)^2+Z^2)^(5/2)For[i=0,i<20,i++,For[j=0,j<15,j++,K[0,j]=0,K[i,0]=0]]For[i=1,i<20,i++, For[j=1,j<15,j++,m[i,j]=NIntegrate[F,x,-10,10,y,-10,10]]]Q=Array[m,18,13]For[i=1,i<19,i++, For[j=1,j<14,j++,K[i,j]=0.25*(m[i,j]+m[(i+1),j]+m[i,(j+1)]+m[(i+1),(j+1)])]]For[i=1,i<19,i++,

Page 3

139

BEAD INTEGRALS.txt For[j=1,j<14,j++,L=L+K[i,j]]]L511.38

F1X=0.1*i-0.1Y=j+11Z=0.125L=0F=(2*(x-X)^2-(y-Y)^2-Z^2)/((x-X)^2+(y-Y)^2+Z^2)^(5/2)For[i=0,i<120,i++,For[j=0,j<11,j++,K[0,j]=0,K[i,0]=0]]For[i=1,i<120,i++, For[j=1,j<11,j++,m[i,j]=NIntegrate[F,x,-10,10,y,-10,10]]]Q=Array[m,118,9]For[i=1,i<119,i++, For[j=1,j<10,j++,K[i,j]=0.25*(m[i,j]+m[(i+1),j]+m[i,(j+1)]+m[(i+1),(j+1)])]]For[i=1,i<119,i++, For[j=1,j<10,j++,L=L+K[i,j]]]L-66.0853

F2X=0.1*i+11.7-0.1Y=j+11Z=0.125L=0F=(2*(x-X)^2-(y-Y)^2-Z^2)/((x-X)^2+(y-Y)^2+Z^2)^(5/2)For[i=0,i<86,i++,For[j=0,j<11,j++,K[0,j]=0,K[i,0]=0]]For[i=1,i<86,i++, For[j=1,j<11,j++,m[i,j]=NIntegrate[F,x,-10,10,y,-10,10]]]Q=Array[m,84,9]For[i=1,i<85,i++, For[j=1,j<10,j++,K[i,j]=0.25*(m[i,j]+m[(i+1),j]+m[i,(j+1)]+m[(i+1),(j+1)])]]For[i=1,i<85,i++, For[j=1,j<10,j++,L=L+K[i,j]]]L19.7699

F3X=0.1*i+11.7-0.1Y=j-1Z=0.125L=0F=(2*(x-X)^2-(y-Y)^2-Z^2)/((x-X)^2+(y-Y)^2+Z^2)^(5/2)For[i=0,i<86,i++,For[j=0,j<11,j++,K[0,j]=0,K[i,0]=0]]For[i=1,i<86,i++, For[j=1,j<15,j++,m[i,j]=NIntegrate[F,x,-10,10,y,-10,10]]]Q=Array[m,84,13]For[i=1,i<85,i++, For[j=1,j<14,j++,K[i,j]=0.25*(m[i,j]+m[(i+1),j]+m[i,(j+1)]+m[(i+1),(j+1)])]]For[i=1,i<85,i++, For[j=1,j<14,j++,L=L+K[i,j]]]L289.328

Page 4


FIT.txt

Chapter 6FIG 6.11

expdata=Import["C:/Experimental2.dat"];expdata=expdata[[Range[2,Length[expdata]]]]pexp=ListPlot[expdata]0.03982,-0.833,0.12542,2.89,0.22659,7.55,0.30441,11.29,0.47561, 18.73,0.60012,24.36,0.75576,30.78,0.98143,40.09,1.24602, 50.36,1.5962,62.06,1.95417,72.16,2.39774,81.73,2.72458, 86.94,3.21484,92.17,3.48721,94.07,3.97747,95.67,4.31209, 95.88,4.72453,95.3,5.06693,94.38,5.48715,92.79,6.242, 89.21,7.03575,84.91,7.83729,80.58,8.59991,76.54,9.36254, 72.71,10.1796,68.88,10.9656,65.55,11.7205,62.55,12.4753, 59.88,13.3858,56.91,14.0784,54.84,14.8643,52.69,15.6192, 50.76,11.7205,62.45,7.8762,80.09,5.70505,91.64,3.89186, 95.59,1.8997,72.18,0.03982,0.85g[Hc_,Hex_,H_]:= Evaluate[x/.Solve[Hc*x*(1-x^2)^(1/2)-H*(1-x^2)^(1/2)+Hex*x\[Equal]0, x][[2]]];V[V0_,Hc_,Hex_,H_]:=V0*g[Hc,Hex,H]*(1-g[Hc,Hex,H]^2)^(1/2)

fitfunc=Sum[Abs[((expdata[[ii,2]]-V[V0,Hc,Hex,expdata[[ii,1]]]))]^2 ,ii,Length[expdata]];FindMinimum[fitfunc,V0,80.,90,Hc,0.00001,0.5,Hex,3.,6., MaxIterations\[Rule]10000]48.1295,V0\[Rule]191.372,Hc\[Rule]0.00142651,Hex\[Rule]4.34506

p1=Plot[V[191.37154365057953`,0.001426513508551305`,4.345056215392623`,H],H, 0,16,PlotRange\[Rule]All,DisplayFunction\[RuleDelayed]Identity];Show[pexp,p1]Export["C:/fit.eps",%]

Page 1

Appendix D

Publications

D.1 List of publications

Published journal papers and conference proceedings. The papers directly related to thesubject of this thesis are included on the following pages.

Papers directly related to the subject of this thesis

• L. Ejsing, M. F. Hansen and A. K. Menon, ”Planar Hall Effect magnetic Sensorfor Micro-Bead Detection”, Conference proceedings: Eurosensors 2003, September2003.

• L. Ejsing, M. F. Hansen, A. K. Menon, H. A. Ferreira, D. L. Graham, P. P. Freitas,”Planar Hall effect sensor for magnetic micro- and nanobead detection”, Appl. Phys.Lett., 84, p. 4729-4731, June 2004.

• L. Ejsing, M. F. Hansen, A. K. Menon, H. A. Ferreira, D. L. Graham, P. P. Fre-itas, ”Magnetic microbead detection using the planar Hall effect”, J. Magn. Magn.Mater., 293, p. 677-684, March 2005.

• R. Marie, S. Schmid, A. Johansson, L. Ejsing, M. Nordstrom, D. Hafliger, C. B. V.Christensen, A. Boisen and M. Dufva, ”Immobilisation of DNA to polymerised SU-8photoresist”, Biosensors and Bioelectronics, 21, p. 1327-1332, March 2005.

Papers not related to the subject of this thesis

• L. Ejsing, K. Smistrup, C. M. Pedersen, N. A. Mortensen and H. Bruus, ”Frequencyresponse in surface-potential driven electrohydrodynamics”, Phys. Rev. E, 73, p.037302 1-4, March 2006.

141

142 APPENDIX D. PUBLICATIONS

Bibliography

[1] Francois Montaigne, Alain Schuhl, Frederic Nguyen Van Dau, and Armando Encinas.Development of magnetoresistive sensors based on planar Hall effect for applicationsto microcompass. Sensors and Actuators A, 81(1-3):324–327, July 2000.

[2] Louise Ejsing, Mikkel F. Hansen, and Aric K. Menon. Planar Hall effect magnetic sen-sor for micro-bead detection. Conference proceedings, Eurosensors 2003, September2003.

[3] Tobias Johansson, Peter Jensen, and Casper Krag. Helmholtz coils for characteri-zation of magnetic sensors. Fagpakkeprojekt 10061, Mikroelektronik Centret (MIC),Technical University of Denmark (DTU), Ørstedsplads, 2800 Kongens Lyngby, April2003.

[4] Louise Ejsing, Mikkel F. Hansen, Aric K. Menon, Hugo A. Ferreira, Daniel L. Graham,and Paulo P. Freitas. Planar Hall effect sensor for magnetic micro- and nanobeaddetection. Applied Physics Letters, 84(23):4729–4731, June 2004.

[5] Louise Ejsing, Mikkel F. Hansen, Aric K. Menon, Hugo A. Ferreira, Daniel L. Graham,and Paulo P. Freitas. Magnetic microbead detection using the planar Hall effect.Journal of Magnetism and Magnetic Materials, 293(1):677–684, May 2005.

[6] Robert C. O’Handley. Modern Magnetic Materials. John Wiley & Sons, Inc., NewYork, 2000.

[7] M. N. Baibich, J. M. Broto, A. Fert, F. Nguyen Van Dau, F. Petroff, P. Etienne,G. Creuzet, A. Friederich, and J. Chazelas. Giant magnetoresistance of (001)Fe/Crmagnetic superlattices. Physical Review Letters, 61(21):2472–2475, November 1988.

[8] Robert L. White. Giant magnetoresistance materials and their potential as read headsensors. IEEE Transactions on Magnetics, 30(2):346–352, March 1994.

[9] H. A. Ferreira, D. L. Graham, P. P. Freitas, and J. M. S. Cabral. Biodetection usingmagnetically labeled biomolecules and arrays of spin valve sensors (invited). Journalof Applied Physics, 93(10):7281–7286, May 2003.

[10] Pierre-A. Besse, Giovanni Boero, Michel Demierre, Vincent Pott, and RadivojePopovic. Detection of a single magnetic microbead using a miniaturized silicon Hallsensor. Applied Physics Letters, 80(22):4199–4201, June 2002.

143

144 BIBLIOGRAPHY

[11] M. A. Parker, T. L. Hylton, K. R. Coffey, and J. K. Howard. Microstructural origin ofgiant magnetoresistance in a new sensor structure based on NiFe/Ag discontinuousmultilayer thin films. Journal of Applied Physics, 75(10):6382–6384, May 1994.

[12] R. L. Edelstein, C. R. Tamanaha, P. E. Sheehan, M. M. Miller, D. R. Baselt, L. J.Whitman, and R. J. Colton. The BARC biosensor applied to the detection of bi-ological warfare agents. Biosensors and Bioelectronics, 14(10–11):805–813, January2000.

[13] M. M. Miller, P. E. Sheehan, R. L. Edelstein, C. R. Tamanaha, L. Zhong, S. Bounnak,L. J. Whitman, and R. J. Colton. A DNA array sensor utilizing magnetic microbeadsand magnetoelectronic detection. Journal of Magnetism and Magnetic Materials,225(1–2):138–144, 2001.

[14] D. K. Wood, K. K. Ni, D. R. Schmidt, and A. N. Cleland. Submicron giant magne-toresistive sensors for biological applications. Sensors and Actuators A, 120(1):1–6,April 2005.

[15] D. L. Graham, H. Ferreira, J. Bernardo, P. P. Freitas, and J. M. S. Cabral. Singlemagnetic microsphere placement and detection on-chip using current line designswith integrated spin valve sensors: Biotechnological applications. Journal of AppliedPhysics, 91(10):7786–7788, May 2002.

[16] D. L. Graham, H. A. Ferreira, P. P. Freitas, and J. M. S. Cabral. High sensitiv-ity detection of molecular recognition using magnetically labelled biomolecules andmagnetoresistive sensors. Biosensors and Bioelectronics, 18(4):483–488, April 2003.

[17] D. L. Graham, H. A. Ferreira, and P. P. Freitas. Magnetoresistive-based biosensorsand biochips. Trends in Biotechnology, 22(9):455–462, September 2004.

[18] D. L. Graham, H. A. Ferreira, N. Feliciano, P. P. Freitas, L. A. Clarke, and M. D.Amaral. Magnetic field-assisted DNA hybridisation and simultaneous detection usingmicron-sized spin-valve sensors and magnetic nanoparticles. Sensors and ActuatorsB, 107(2):936–944, June 2005.

[19] Guanxiong Li, Vikram Joshi, Robert L. White, Shan X. Wang, Jennifer T. Kemp,Chris Webb, Ronald W. Davis, and Shoueng Sun. Detection of single micron-sizedmagnetic bead and magnetic nanoparticles using spin valve sensors for biologicalapplications. Journal of Applied Physics, 93(10):7557–7559, May 2003.

[20] G. Mihajlovic, P. Xiong, S. von Molnar, K. Ohtani, H. Ohno, M. Field, and G. J.Sullivan. Detection of single magnetic bead for biological applications using an InAsquantum-well micro-Hall sensor. Applied Physics Letters, 87(11):112502–1–3, Septem-ber 2005.

[21] J. B. Johnson. Thermal agitation of electricity. Bell Laboratories Record, 3(6):185–187, February 1927.

BIBLIOGRAPHY 145

[22] J. B. Johnson. Thermal agitation of electricity in conductors. Physical Review,32(1):97–109, July 1928.

[23] J. Wilson and J. Hawkes. Optoelectronics an introduction. Prentice Hall, Hertford-shire, third edition, 1998.

[24] B. Raquet. Electronic noise in magnetic materials and devices. Laboratoire dePhysique de la Matiere Condensee de Toulouse, LPMCT-LNCMP-INSA, France.

[25] J. Briaire, L. K. J. Vandamme, and M. A. M. Gijs. Resistance and 1/f noise de-pendence on magnetic field in single Ni80Fe20 layers and Ni80Fe20/Cu multilayers.Proceedings of the 14th International Conference on Noise in Physical Systems and1/f Fluctuations, 1997.

[26] P. P. Freitas, H. A. Ferreira, D. L. Graham, L. A. Clarke, M. D. Amaral, V. Martins,L. Fonseca, and J. S. Cabral. in M. Johnson (Ed.) Magnetoelectronics, chapter 7:Magnetoresistive DNA chips, pages 331–373. Elsevier Academic Press, Oxford, firstedition, December 2004.

[27] J. Nogues and I. K. Schuller. Exchange bias. Journal of Magnetism and MagneticMaterials, 192(2):203–232, February 1999.

[28] Richard M. Bozorth. Ferromagnetism. IEEE press, Wiley-Interscience, John Wiley& Sons, Inc., Hoboken, New Jersey, 2003.

[29] N. F. Mott. The resistance and thermoelectric properties of the transition metals.Proceedings of the Royal Society of London. Series A, Mathematical and PhysicalSciences, 156(888):368–382, August 1936.

[30] V. D. Ky. Planar Hall Effect in Ferromagnetic Films. Physica Status Solidi,26(2):565–569, 1968.

[31] Y. Shirakawa. On the Longitudinal Magneto-Resistance Effect at VariousTemperatures in Iron-Nickel Alloys. Tohoku Imperial University – Science Reports,27(4):485–531, January 1939.

[32] David J. Griffiths. Introduction to electrodynamics. Prentice-Hall, Inc., New Jersey,second edition, 1989.

[33] Veronique Gehanno, Paulo P. Freitas, Anabela Veloso, Joao Ferreira, BernadoAlmeida, J. B. Sousa, A. Kling, J. C. Soares, and M. F. da Silva. Ion beam de-position of Mn-Ir spin valves. IEEE Transactions on Magnetics, 35(5):4361–4367,September 1999.

[34] Rodolphe Marie, Silvan Schmid, Alicia Johansson, Louise Ejsing, Maria Nordstrom,Daniel Hafliger, Claus BV Christensen, Anja Boisen, and Martin Dufva. Immobil-isation of DNA to polymerised SU-8 photoresist. Biosensors and Bioelectronics,21(7):1327–1332, March 2006.

146 BIBLIOGRAPHY

[35] Dirch Hjorth Petersen and Peter Morten Moselund. Simulation and characterizationof a Hall sensor with automated testing. Special course, Department of Micro andNanotechnology (MIC), Technical University of Denmark (DTU), Ørstedsplads, 2800Kongens Lyngby, July 2003.

[36] Stanford Research Systems. Model SR830 DSP Lock-in Amplifier. Manual, Novem-ber 2002.

[37] Alexandre Pesci. LEMO Unipole & Multipole Connectors. Catalogue, April 2000.

[38] G. Glikmann, C. H. Mordhorst, and C. Koch. Monoclonal antibodies for the directdetection of influenza-A virus by ELISA in clinical specimens from patients withrespiratory infections. Clinical and Diagnostic Virology, 3(4):361–369, 1995.

[39] H. A. Ferreira, N. Feliciano, D. L. Graham, and P. P. Freitas. Effect of spin-valvesensor magnetostatic fields on nanobead detection for biochip applications. Journalof Applied Physics, 97(10):10Q904,1–3, May 2005.

[40] Bangs Laboratories. Technote 201: Working with microspheres. Technical Report201, Bangs Laboratories Incorporate, www.bangslabs.com, October 2002.

Planar Hall Effect Magnetic Sensor for Micro-Bead Detection

L. Ejsing, M. F. Hansen and A. K. Menon

Mikroelektronik Centret (MIC), Technical University of Denmark (DTU), Building 345 East, 2800 Kgs. Lyngby, Denmark

e-mail: [email protected] http://www.mic.dtu.dk

Summary: Magnetic sensors based on the planar Hall effect (PHE) of Ni have been fabricated and characterized. The response to an applied magnetic field has been measured without and with a coating of commercially available 2.8 µm magnetic beads for bioseparation (Dynabeads M-280, Dynal, Norway). It is demonstrated that the technique is capable of detecting just a few magnetic beads, i.e., that the technique is feasible for magnetic biosensors. Keywords: Planar Hall effect, magnetic bead detection Category: 3 (Magnetic physical devices)

1 Introduction Planar Hall effect sensors have been used to detect nanotesla magnetic fields [1,2]. Also, the exchange anisotropy behavior can be characterized using thin films’ planar Hall effect [3]. We demonstrate that a planar Hall effect sensor can be used to detect magnetic beads used for bioapplications.

The anisotropic magnetoresistance (AMR) effect is present in ferromagnetic materials. It results in a dependency of the resistivity on the orientation of the material’s magnetization with respect to the current through the material. The electrical output signal in the planar Hall geometry (Fig. 1) is [4]

( ) ( )φφρρ

cossin)(

t

IV

x

y

⊥−=

where t is the thickness of the metal layer, Ix is the applied current, and φ is the angle between the current and the in-plane magnetization vector. ρ|| and ρ⊥ are the resistivities when the magnetization is parallel and perpendicular to the current, respecti-

easy axis φ

V

I

y x

M r

Fig. 1. Planar Hall sensor geometry: The magnetic field to be detected is in the detection plane, applied in the y-direction. In this plane, the current is applied in the x-direction and a voltage is measured in the y-direction.

vely. The difference between ρ|| and ρ⊥ is respon-sible for the anisotropic magnetoresistance and the planar Hall effect. If they were equal, which is the common case, there would be no Vy response to a planar applied magnetic field, i.e., a field applied in the y-direction. For Ni the resistivity variation is approximately 2%.

A schematic drawing of the planar Hall sensor is presented in Fig. 1. Initially the magnetization vec-tor lies along the easy axis and the current direction, in which Ix = 250 µA. A magnetic field applied in the y-direction induces a rotation of the magnetiza-tion by an angle φ. This changes the electrical out-put signal by the amount Vy.

The planar Hall geometry is ideal for measuring magnetized beads because these have a significant in-plane field component when magnetized parallel to the sensor plane.

If the easy axis lies in the x-direction and an external field, By, is applied in the y-direction, a first approximation of the energy of a magnetisation vector at an angle φ to the x-direction is

( ) ( )φφ sinsin 2yan MVBEE −=

where Ean is the anisotropy energy, and V is the volume. At equilibrium

( )an

y

an

y

B

B

E

VMB≡=

2sin φ

where Ban is the anisotropy field. For small angles, sin(φ)cos(φ) reduces to sin(φ), and the signal to

an

yx

y B

B

t

IV

)( ⊥−=

ρρ for any BB ≤

Therefore, in small applied fields with respect to the anisotropy field of the easy axis (typical values for our sensors are Ban= 2-3 mT) the response of the sensor is linear.

1.1 Dynabead signal

The low-field magnetic volume susceptibility of the M-280 Dynabeads is χ = 0.13 [SI] [5]. From this

the low-field magnetic moment of a Dynabead of 2.8 µm in diameter can be estimated as m = χHV, where H is the applied field intensity, and V= π(2.8µm)3/6 is the volume of the Dynabead.

The expression for a dipole’s magnetic field in spherical coordinates is [6]

( ) ( )( )θθθπ

µ ˆsinˆcos24 3

0 += rrm

Bdip

r

where r is the distance from the dipole to the observation point, θ is the angle between the dipole moment and the observation point, and µ0 is the vacuum permeability.

The bead is placed on top of the sensor with its magnetic moment aligned in the y-direction by the external field. Hence the dipole field from the bead will have the opposite direction just below the bead, i.e., in the –y-direction. Because the magnetic field from the bead has the opposite direction of the applied external field, the presence of the beads will lead to a reduction of the magnetic field felt by the sensor. This change in sensitivity to an applied external magnetic field forms the basis of the bead sensing principle.

Assuming dipole behavior of the bead, and only taking the point directly under the dipole moment into account, the angle is θ=90°, and the magnitude of the magnetic flux density from the bead at the sensor reduces to

30

4~

rVH

Bbeadχ

πµ

−

The distance from the magnetic moment to the observation point is, in this case, the distance from the center of the bead to the sensor surface. For a bead placed on the sensor surface r ≈ 1.4 µm.

It should be noted that assuming dipole behaviour of a Dynabead is a very crude model. The bead consists of a large number (105-106) of superpara-magnetic nanoparticles embedded in a polymer, and though these might behave as dipoles the sum is likely to be more complicated. Another problem is that the field produced by the bead is considered homogeneous at the sensor surface, which is only true for one dipole viewed from exactly θ=90°. The purpose of this approach is to give an estimate of the order of magnitude, only.

2 Experiments and results The sensors are fabricated using conventional litho-graphy. Crosses are patterned on a resist coated silicon wafer, and 20 nm of Ni is deposited in an e-beam evaporation system. During deposition a uniform field of 8 mT is applied in order to ensure formation of an easy magnetic axis. A 2 nm thick gold cap is deposited on top in order to protect the nickel from corrosion. Finally, the resist is lifted off with the excess of metal to yield the PHE sensors. Bonding pads are made of gold using the same procedure.

Initially, the magnetization is set along one of the easy axes. When the magnetization is rotated away from the easy axis by a field applied in the y-direction, the change in magnetization produces an electrical output signal.

The output signal is measured as a function of applied field, µ0Hext, between 0 and 0.6 mT, first for the sensor without magnetic beads, then for the sensor with magnetic particles placed on top of the sensor (Fig. 2). The response change is due to the presence of beads.

Fig. 2: Sensor area (20µm × 20µm) with 2.8 µm Dyna-beads on top. To the right a large cluster of beads is positioned away from the sensing region. The black is sedimentary dirt from the Dynabeads’ solution.

The sensor area with clusters of beads on top is

shown in Fig. 2. The sample depicted was prepared in the following way. A droplet containing beads was placed on top of the sample with one sensor using a syringe. Then the sample was left to dry and the beads to settle on the surface. The black is sedimentary dirt from the Dynabeads’ solution. The dirt covers a large cluster of beads positioned to the right of the sensing area.

In Fig. 2 the black dotted line shows the sensitive area of the sensor. Within this line, beads contribute to the signal with full magnitude. When counting, we get N=6 in this area. The white dotted lines show the detection limit for a single bead estimated from the uncertainty on the measurements. Outside this square the contribution from a single bead is less than the uncertainty on the slope. Between the black and the white line the signal from a particle decreases as ~1/r3. Within the white square, N=31.

Calibration of the sensor in homogeneous applied fields up to 0.6 mT is conducted before placing beads on top of the sensor. The calibration yields a linear field response and the sensitivity S = 7.0 µV/mT. At 0.6 mT the dipole field at the sensor from one bead is Bbead ~ 0.026 mT and the expected

signal change is Vbead ~ -0.18 µV. Thus to produce 1 µV change in the signal ~6 Dynabeads are needed.

The results of the output voltage as a function of applied magnetic field for the sensor with and without beads on top of the sensitive area are shown in Fig. 3. The distinction between the two types of response is clear, which indicates the potential for biosensor applications of this type of sensor.

When the beads are magnetized, the bead moments align with the applied field. Hence, the effective magnetic field at the sensor, Heff, and the output signal will be reduced according to:

−= 300 41~

rVN

SHSHV exteffy πχ

µµ

where N is the number of beads positioned on the sensor. An estimate of N from the slopes found in Fig. 3 according to this simple analysis leads to N ~ 12. It should be noted that the sensor is most sensitive to beads situated on the sensor, but also that clusters of beads in the vicinity of the sensor can contribute.

0,0 0,1 0,2 0,3 0,4 0,5 0,60

1

2

3

4

5

slope=3.5µV/mT

slope=7.0µV/mT

Sig

nal (

µV)

Applied field (mT)

no beads with beads (meas. 1) with beads (meas. 2)

Fig. 3. Offset corrected sensor output signal as a function of applied magnetic field with and without beads. The measurements with beads are conducted twice to ensure repeatability.

After the first series of measurements, the beads

were washed away with water and a droplet of diluted beads was placed on top of the sensor. The new sample is shown in Fig. 4. In this picture damages from the first experiments are visible along with the coating of just a few beads on top of the sensing area. Only two beads are situated within the sensitive area marked by black dotted lines. A third bead has the possibility of contributing to the signal; it is placed within the white dotted lines.

Fig. 4: Sensor area (20µm × 20µm) with a coating of less 2.8 µm Dynabeads on top compared to Fig. 2. Only two beads are situated at the sensing region.

Before placing the new coating of beads on top of

the sensor, the sensor was calibrated again, and the sensitivity of S = 7.0 µV/mT was obtained once more.

The sensor output signal corresponding to this coating of beads is shown in Fig. 5. The black squares are measurements without any beads on top of the sensor, the open and filled circles are the measurements corresponding to the bead coverage shown in Fig. 2, and the open triangles are the measurements corresponding to the bead coverage shown in Fig. 4.

0,0 0,1 0,2 0,3 0,4 0,5 0,60

1

2

3

4

5

slope=6.8µV/mT

Sig

nal (

µV)

Applied field (mT)

no beads with beads (meas. 1) with beads (meas. 2) with few beads

Fig. 5. Offset corrected sensor output signal as a function of applied magnetic field with and without beads. The open triangles (few beads) and the open+solid circles (more beads) correspond to the bead coverage shown in Fig. 4 and 2, respectively.

When estimating the number of beads contri-buting to the signal from the slopes in Fig. 5, we get N~2, which is also observed in the picture (Fig. 4). Remembering the very crude approximations made to give an order of magnitude estimate of the

magnetic field from the bead, the simple estimate works surprisingly well.

Conclusion It has been demonstrated that the planar Hall effect can be used for detecting commercial micrometer-sized magnetic beads used for bioapplications. The number of detected beads is in fine agreement with simple theoretical estimates. The PHE sensor is sensitive enough to detect just a few beads. Thus, the planar Hall effect has potential for use in magnetic biosensors, e.g., for DNA detection. An advantage of the planar Hall effect sensor compared to giant magnetoresistive (GMR) sensors is the simple fabrication scheme, which reduces the fabrication cost.

Future work includes future theoretical work, sensor optimization and implementation with micro-fluidic systems. The sensor will furthermore be demonstrated as a biosensor where DNA fragments bind the beads to the surface.

References [1] F. Montaigne, A. Schuhl, F. N. Van Dau, A. Encinas, Sensors and Actuators 81 (2000) 324-327.

[2] F. N. Van Dau, A. Schuhl, J. R. Childress, M. Sussiau, Sensors and Actuators A 53 (1996) 256-260.

[3] Z. Q. Lu, G. Pan, W. Y. Lai, J. Appl. Phys. 90 (2001) 1414-1418.

[4] R. C. O’Handley. Modern Magnetic Materials, John Wiley & Sons, Inc. (2000).

[5] http://www.dynal.no

[6] D. J. Griffiths. Introduction to electrodynamics, Prentice-Hall, Inc. (1989).

APPLIED PHYSICS LETTERS VOLUME 84, NUMBER 23 7 JUNE 2004

Planar Hall effect sensor for magnetic micro- and nanobead detectionL. Ejsing,a) M. F. Hansen, and A. K. MenonDepartment of Micro and Nanotechnology (MIC), Technical University of Denmark (DTU),Building 345 East, DK-2800 Kongens Lyngby, Denmark

H. A. Ferreira, D. L. Graham, and P. P. FreitasInstitute of Engineering of Systems and Computers—Microsystems and Nanotechnologies (INESC-MN),Rua Alves Redol 9, Lisbon 1000-029, Portugal

~Received 26 January 2004; accepted 8 April 2004; published online 20 May 2004!

Magnetic bead sensors based on the planar Hall effect in thin films of exchange-biased permalloyhave been fabricated and characterized. Typical sensitivities are 3mV/Oe mA. The sensor responseto an applied magnetic field has been measured without and with coatings of commercially available2 mm and 250 nm magnetic beads used for bioapplications~Micromer-M and Nanomag-D,Micromod, Germany!. Detection of both types of beads and single bead detection of 2mm beads isdemonstrated, i.e., the technique is feasible for magnetic biosensors. Single 2mm beads yield 300nV signals at 10 mA and 15 Oe applied field. ©2004 American Institute of Physics.@DOI: 10.1063/1.1759380#

a

mgete.mea

ory

etoesod.groatgha

,a

en

e

reeslie

theas

pytion,usethe

all

of

de-

ruc-

Detection of magnetic beads for bioapplications hbeen carried out with giant magnetoresistance~GMR!sensors,1–3 spin valve sensors,4–7 and a silicon Hall sensor.8

Planar Hall sensors have been used to detect nanoteslanetic fields.9,10 This letter demonstrates that an exchanbiased permalloy planar Hall sensor can be used to demicro- and nanomagnetic beads used for bioapplications

Formerly, a planar Hall sensor made of nickel was deonstrated capable of detecting superparamagnmicrobeads.11 Here, we report on the use of permalloy assensing material since it has a higher anisotropic magnetsistance~AMR! effect than nickel. To control the anisotropand to achieve a well-defined single domain initial magnzation state, the permalloy layer is exchange coupled tMnIr antiferromagnetic layer. The 200 Å thick exchangcoupled permalloy layer has an increased effective aniropy field,Han'59 Oe, defining the sensor saturation fiel

The planar Hall effect is based on the AMR of ferromanetic materials. The transverse voltage on a planar Hall cdepends on the orientation of the magnetization of the mrial with respect to the longitudinal current running throuthe material. The electrical output signal in the planar Hgeometry~inset of Fig. 1! is12

Vy5~r i2r'!I x

tsin~f!cos~f!, ~1!

wheret is the metal layer thickness,I x is the applied currentandf is the angle between the current and the in-plane mnetization vector,M . r i andr' are the resistivities when thmagnetization is parallel and perpendicular to the currerespectively.Dr5(r i2r') is responsible for AMR and theplanar Hall effect. For a 200 Å thick permalloy film, thresistivity variation is approximately 2.2%.13

A schematic drawing of the planar Hall sensor is psented in the inset of Fig. 1. Initially, the magnetization lialong the easy axis, which is also the direction of the app

a!Electronic mail: [email protected]

4720003-6951/2004/84(23)/4729/3/$22.00Downloaded 07 Dec 2004 to 192.38.67.112. Redistribution subject to AIP

s

ag--ct

-tic

e-

i-a

-t-

-sse-

ll

g-

t,

-

d

current,I x . When a magnetic field,Hy , is applied in theydirection, the magnetization rotates by an anglef in thesensor plane. This changes the electrical output signal byamountVy . For small angles, the signal can be expressed

Vy5~r i2r'!l x

t

Hy

Han. ~2!

Therefore, in small applied fields compared to the anisotrofield, the response of the sensor is linear. For bead detecHy in expression~2! represents the applied external field plthe sum of they components of the field created by thhomogeneously magnetized spherical beads weighed bybead-to-sensor area fraction.

A micrograph of the cleanroom fabricated planar Hsensor is shown in Fig. 1. The sensor layer structure Ta~30Å!/NiFe ~50 Å!/MnIr ~200 Å!/NiFe ~200 Å!/Ta ~30 Å! ~see

FIG. 1. Micrograph of the planar Hall sensor. The cross is madeexchange-biased permalloy, and the central area is the 10mm310mm sen-sitive area of the sensor. Current leads are made of 0.3mm thick Al. The topinset shows the planar Hall sensor geometry. The magnetic field to betected is applied in the detection plane, along they direction. In this plane,the current is applied in thex direction and the voltage,Vy , is measured inthe y direction. The bottom inset illustrates the cross-sectional layer stture of the sensor.

9 © 2004 American Institute of Physics license or copyright, see http://apl.aip.org/apl/copyright.jsp

http://dx.doi.org/10.1063/1.1759380

ologies

4730 Appl. Phys. Lett., Vol. 84, No. 23, 7 June 2004 Ejsing et al.

Downloaded 07 De

TABLE I. Physical properties of Micromod-M beads and Nanomag-D beads.a

Diameter Concentration Density xb

Micromer-M 2 mm .25 mg/ml 1.4 g/cm3 0.360.1 ~SI!Nanomag-D 250 nm .10 mg/ml 4.0 g/cm3 662 ~SI!

aSee Ref. 14.bMeasured at Institute of Engineering of Systems and Computers–Microsystems and Nanotechn~INESC–MN!.

he

Oag

amrotee

ethpie;

se-

in

Iag

t

re

ip

en, atnal

re-thehes

iversorpileds noo be

de-

Aye isingads

esn

a

oilshewereberscifiedcific

ally,of

ringult is

cross-sectional diagram in the inset of Fig. 1! was preparedby ion-beam deposition on a passivated 3 in. Si wafer, tcovered by sputtered antireflecting TiWN~150 Å!. Ion-beamdeposition conditions are given in Gehannoet al.13 Here,MnIr stands for Mn76Ir24 and NiFe stands for Ni80Fe20. Dur-ing ion-beam metal deposition, a magnetic field of 40was applied in order to form an easy axis in the ferromnetic NiFe layers. Sensors~crosses of 10mm310mm) werepatterned by direct write laser lithography and ion-bemilling. Al current leads~0.3 mm thick! were sputtered aftea soft etch of the contact surface and defined by photolithraphy and lift off. The sensor structure was then passivaby a sputtered SiO2 ~2000 Å! layer and contact pads weropened by reactive ion etching.

The active sensing layer is the NiFe~200 Å! layer. Forcharacterization of the planar Hall sensors, electromagncoils were used to provide a tuneable magnetic field andsensor response signal was measured as a function of apfield strength. Direct currents of 1, 5, and 10 mA are appland direct voltage drops are measured using a voltmeterelectronic noise reduction is used. A typical sensitivity isS53 mV/Oe mA in the range of615 Oe, where the signal ilinear within 2.8%. This sensitivity is reproducible for a spcific sensor and for sensors on the same chip.

For demonstration of bead detection, a droplet containg superparamagnetic Micromod beads~2 mm or 250 nmdiameter,14 physical properties are summarized in Table!was placed on top of the sensor while measuring the voltdrop. The beads were magnetized with a215 Oe in-planeapplied field generated by electromagnetic coils leadingestimated local dipole fields just below the beads of11.5 Oefor 2 mm beads, and130 Oe for 250 nm beads.

Figure 2 presents the bead detection results. Curthrough the sensor is 5 mA and the applied field is215 Oe.At time t5100 s, the 2mm beads are added onto the ch

FIG. 2. Bead detection measurements. A constant voltage (V052464mV) is subtracted from the signal to give a baseline at 0mV. At timest5100 s andt5700 s the 2mm beads are added to the chip, and at timt5300 s andt51050 s they are washed away. The same is done for 250beads ~added att51400 s andt52250 s, removed att51650 s andt52650 s). The signal rises when beads settle on top of the sensorreturns to the baseline when the beads are removed from the sensor.

c 2004 to 192.38.67.112. Redistribution subject to AIP

n

e-

g-d

tice

lieddno

-

e

o

nt

and the signal rises as the beads settle on the sensor. Thtime t5300 s, the beads are washed away and the sigreturns to the baseline. At timest5700 s andt51050 s, re-spectively, the experiment is repeated yielding the samesult. At time t51400 s, the 250 nm beads are added tochip and at timet51650 s the beads are washed away. Texperiment is repeated with the 250 nm beads at timet52250 s andt52650 s, respectively. The 250 nm beads ghigher saturation signals than the 2mm beads due to theihigher susceptibility and higher number on top of the senarea. Saturation of the signal occurs when the beads areup on top of the sensor and the addition of another bead ilonger sensed because it is too far away from the sensor tdetected.

Figure 3 shows sensor response to an increasing orcreasing number of beads over the sensing area~2 mmMicromod-M beads!. Current through the sensor is 10 mand applied field is215 Oe. The estimate of how manbeads are floating over the sensitive area at a given timcarried out by visual inspection through a microscope durthe measurements. Specifically, the large numbers of be

m

nd

FIG. 3. Detection of clusters. The field applied with electromagnetic cwas Happ5215 Oe. ~a! The sensor signal measured while observing tsensor through an optical microscope keeping track of when beadsfloating over the sensor. The numbers written on the chart are the numof beads entering or leaving the space over the sensitive area at the spetime. The estimate of bead number giving the voltage change at a spetime was made by visual inspection through the microscope. Specificthe large numbers are connected with uncertainty due to the difficultycounting beads in a limited amount of time.~b! Summary of the data fromchart~a!. The immediate change in signal for a given bead number enteor leaving the space over the sensor. For low bead numbers, the resVbead;0.3mV.
license or copyright, see http://apl.aip.org/apl/copyright.jsp

gelo

i

edsoehg

fe

al

si

osioit

oouth

ze

zsonithe

e is0ith

-r as

d R.

R.

ng,

bral,

ppl.

ens.

.

ys.

tors

tua-

-

usa,

4731Appl. Phys. Lett., Vol. 84, No. 23, 7 June 2004 Ejsing et al.

are connected with uncertainty since it is difficult to judthe exact number of beads constituting a cluster. Forbead numbers, the resulting one-bead signal isVbead

;0.3mV. The noise level for these dc measurements200–300 nV.

In conclusion, detection of commercially available 2mmand 250 nm superparamagnetic beads~Micromer-M andNanomag-D!14 used for bioapplications was demonstratusing an exchange-biased permalloy planar Hall senHence, this planar Hall sensor has potential as a magnbiosensor, e.g., for deoxyribonucleic acid detection. Ttechnique is sufficiently sensitive to detect only a few manetic beads and therefore also biomolecules even if onlyare present in the sample. For small numbers of 2mm beads,bead detection experiments showed a single-bead signVbead;0.3mV.

Due to the cross geometry, the planar Hall sensor useits active surface for bead/biomolecule detection, whichnot the case for meandering-type GMR or spin valve senswhere almost one-half of the target biomolecules willover a nonsensing area. A second advantage, not explhere, concerns the theoretically higher signal-to-noise~S/N!ratio for planar Hall sensors when compared with GMRspin valve sensors. For comparable sensing areas, alththe spin valve sensor yields signals 5 to 10 times higher,planar Hall sensor should have a noise level~in the 1/f domi-nated low-frequency regime! about 20 times lower offeringthe possibility of single bead detection for nanometer-sibeads.

Using a lock-in amplifier to obtain a bandwidth of 1 Hwill improve the S/N considerably. For the reported sensize with an applied current of 15 mA and defining the mimum detectable signal as four times the noise level, theoretical 1/f noise level is'0.6 nV and the sensor should b

Downloaded 07 Dec 2004 to 192.38.67.112. Redistribution subject to AIP

w

s

r.tice-w

of

allsrs,ted

rghe

d

r-e-

able to detect a single 250 nm bead. If the sensor sizreduced to 2.5mm32.5mm and the current is reduced to 1mA, it should be possible to detect a single 40 nm bead wx56 ~SI!.

Future work includes integration with microfluidic systems along with a demonstration of the planar Hall sensoa biosensor and single micro- and nanobead detection.

1D. R. Baselt, G. U. Lee, M. Natesan, S. W. Metzger, P. E. Sheehan, anJ. Colton, Biosens. Bioelectron.13, 731 ~1998!.

2R. L. Edelstein, C. R. Tamanaha, P. E. Sheehan, M. M. Miller, D.Baselt, L. J. Whitman, and R. J. Colton, Biosens. Bioelectron.14, 805~2000!.

3M. M. Miller, P. E. Sheehan, R. L. Edelstein, C. R. Tamanaha, L. ZhoS. Bounnak, L. J. Whitman, and R. J. Colton, J. Magn. Magn. Mater.225,138 ~2001!.

4D. L. Graham, H. Ferreira, J. Bernardo, P. P. Freitas, and J. M. S. CaJ. Appl. Phys.91, 7786~2002!.

5H. A. Ferreira, D. L. Graham, P. P. Freitas, and J. M. S. Cabral, J. APhys.93, 7281~2003!.

6D. L. Graham, H. A. Ferreira, P. P. Freitas, and J. M. S. Cabral, BiosBioelectron.18, 483 ~2003!.

7G. Li, V. Joshi, R. L. White, S. X. Wang, J. T. Kemp, C. Webb, R. WDavis, and S. Sun, J. Appl. Phys.93, 7557~2003!.

8P. A. Besse, G. Boero, M. Demierre, V. Pott, and R. Popovic, Appl. PhLett. 80, 4199~2002!.

9F. Montaigne, A. Schuhl, F. N. Van Dau, and A. Encinas, Sens. Actua81, 324 ~2000!.

10F. N. Van Dau, A. Schuhl, R. J. Childress, and M. Sussiau, Sens. Actors, A53, 256 ~1996!.

11L. Ejsing, M. F. Hansen, and A. K. Menon,Proceedings of the 17th European Conference on Solid-State Transducers, Eurosensors 2003, 21–24September 2003, Guimaraes, Portugal~University of Minho, Guimara˜es,Portugal, 2003!, p. 1095.

12R. C. O’Handley,Modern Magnetic Materials~Wiley, New York, 2000!.13V. Gehanno, P. P. Freitas, A. Veloso, J. Ferreira, B. Almeida, J. B. So

A. Kling, J. C. Soares, and M. F. Silva, IEEE Trans. Magn.35, 4361~1999!.

14http://www.micromod.de

license or copyright, see http://apl.aip.org/apl/copyright.jsp

ARTICLE IN PRESS

Journal of Magnetism and Magnetic Materials 293 (2005) 677–684

0304-8853/$

doi:10.1016

Correspfax: +4545

E-mail a

www.elsevier.com/locate/jmmm

Magnetic microbead detection using the planar Hall effect

Louise Ejsinga,, Mikkel F. Hansena, Aric K. Menona, Hugo A. Ferreirab, DanielL. Grahamb, Paulo P. Freitasb

aDepartment of Micro and Nanotechnology (MIC), Technical University of Denmark (DTU), Building 345 east,

DK-2800 Kongens Lyngby, DenmarkbInstitute of Engineering of Systems and Computers—Microsystems and Nanotechnologies (INESC-MN), Rua Alves Redol 9,

Lisbon 1000-029, Portugal

Available online 29 March 2005

Abstract

Magnetic sensors based on the planar Hall effect of exchanged-biased permalloy have been fabricated and

characterized. It is demonstrated that the sensors are feasible for detecting just a few commercial 2.0 mmmagnetic beads

commonly used for bioseparation (Micromer-M, Micromod, Germany) and that the sensor sense current is sufficient to

generate a signal from the beads.

r 2005 Elsevier B.V. All rights reserved.

PACS: 75; 47; m

Keywords: Hall effect (planar); Micromer-M; Magnetic sensor; Biosensor; Permalloy (Ni80Fe20); MnIr (Mn76Ir24); Detection of

microspheres

1. Introduction

In the lab-on-a-chip concept all steps in analyz-ing a sample are assembled on a single chip. Theidea is to use such a chip directly after taking, e.g.,a blood sample to obtain an immediate answer asto the presence of a specific molecule. Our focushas been on the detection step, where we aim to

- see front matter r 2005 Elsevier B.V. All rights reserve

/j.jmmm.2005.02.071

onding author. Tel.: +454525 6336;

88 7762.

ddress: [email protected] (L. Ejsing).

use magnetic beads and integrate bead detectionwith microfluidic channels.Detection of superparamagnetic beads for bio-

sensor applications has been demonstrated usingGMR sensors [1–3], spin valve sensors [4–7], and asilicon Hall sensor [8]. Planar Hall effect sensorshave been shown to exhibit nano-Tesla sensitivity[9,10] and have recently been investigated for theiruse in magnetic biodetection [11]. We propose theuse of microfabricated planar Hall effect (PHE)sensors and demonstrate bead detection in zeroexternally applied field using an exchange-biasedpermalloy sensor. This sensor is significantly

d.

www.elsevier.com/locate/jmmm

ARTICLE IN PRESS

Ix

Vy

x

y

M

easy axis

φ

Fig. 1. Planar Hall geometry. A current is applied in the x-

direction and a voltage is measured in the y-direction. The

magnetization vector,M, lies in the x–y-plane at an angle, f; to

L. Ejsing et al. / Journal of Magnetism and Magnetic Materials 293 (2005) 677–684678

improved compared to the unbiased nickel PHEsensor previously demonstrated [11].The purpose of exchange-biasing the sensor is to

ensure a sufficient uniaxial anisotropy, a well-defined single domain state and to introduce aunidirectional anisotropy. The sign of the sensoroutput depends upon whether the film magnetiza-tion is initially parallel or antiparallel to the sensorcurrent (Section 2) and thus it is important to havea well-defined initial orientation. For the pre-viously demonstrated Ni sensor, which exhibiteduniaxial anisotropy, this was achieved by saturat-ing the sensor along one of the easy directionsprior to each measurement. Due to the unidirec-tional anisotropy this step is now eliminated and,additionally, the higher anisotropic magnetoresis-tance of permalloy compared to Ni results in alarger signal.

the current direction. The easy axis is along the current

direction.

2. Planar Hall effect sensor principle

The magnetic sensor is based on the anisotropicmagnetoresistance of ferromagnetic materials. Forsuch materials, Ohm’s law can be written [12]

E!

¼ MðJ MÞ½rjj r? þ r?Jþ rHM J, (1)

where M denotes a unit vector along the magne-tization direction and J is the current density. rH isthe ordinary Hall resistivity arising from theLorentz force deflection of the charge carriers. rjjand r? are the resistivities when the magnetizationvector is parallel and perpendicular to the currentdensity, respectively. Generally, rjj4r? and forNi–Fe alloys Dr r1av ðrjj r?Þ r

1av attains

values up to 3% at room temperature [13]. Asensor based on the planar Hall geometry illu-strated in Fig. 1 can be used for detection of smallmagnetic fields [9–11,14]. The sensor consists of athin ferromagnetic film through which a current Ix

is applied in the x-direction and the voltage Vy ismeasured in the y-direction. The magnetizationvector lies in the plane of the sensor at an angle fto the current direction. For this geometry Eq. (1)reduces to

Ey ¼ ðrjj r?ÞJx sinf cosf (2)

and the measured voltage becomes

Vy ¼ 12IxDR sinð2fÞ, (3)

where DR ðrjj r?Þ t1film and tfilm is the filmthickness. The basic principle of the sensor is thefollowing. The easy axis and the easy direction ofmagnetization are defined in the film along thecurrent direction by exchange-coupling to anantiferromagnetic material. If a magnetic field issubsequently applied perpendicular to the easyaxis the magnetization vector rotates away fromthe easy direction, giving rise to an electricalresponse according to Eq. (3). To model the sensorresponse, we can to a first approximation write themagnetic energy density of the film as

U ¼ Kusin2 f UE cosf m0MHy sinf, (4)

where Ku is the uniaxial anisotropy constant, UE

is the exchange energy density, and Hy is themagnetic field affecting the sensor along the y-direction. For small angles, cos f 1; and mini-mization of the energy yields

sinf m0MHy

2Ku þ UE

Hy

HC þ HE, (5)

ARTICLE IN PRESS

P

m

r

H

bead

z

sensor surface

Fig. 3. Magnetic bead with moment, m, positioned relative to

the sensor surface. The bead is magnetized in the field, H, and

produces a field in the opposite direction at the sensor.

L. Ejsing et al. / Journal of Magnetism and Magnetic Materials 293 (2005) 677–684 679

which is valid for Hy5HC þ HE: HC is theanisotropy field, and HE is the exchange couplingfield. Insertion in Eq. (3) yields the ideal sensorresponse

V y IxDRHy

HC þ HE. (6)

The expected sensor response, plotted in Fig. 2, islinear to within 2% for jHyðHC þ HEÞ

1jo0:2

with the sensitivity

S0 V y

HyIx

¼DR

HC þ HE. (7)

The PHE magnetic sensor is sensitive to magneticfields in the sensor plane. This is utilized to detectthe dipole field from magnetic beads as illustratedin Fig. 3. The applied magnetic field magnetizesthe bead and creates a dipole field, which isantiparallel to the applied field in the sensor plane.Thus, the presence of magnetic beads gives rise toa reduction of the applied field, which can bedetected.The magnetic beads contain superparamagnetic

nanoparticle inclusions and will show a linearresponse at low fields. Hence, the bead magnetiza-tion can be written M ¼ wH; where w is the beadsusceptibility (including demagnetization effects)and H is the field intensity containing contribu-tions from the applied field, the field generated by

-1.0

-0.8

-0.6

-0.4

-0.2

0

0.2

0.6

0.4

0.8

1.0

-1.0 1.0-0.5 0.50

exact

linear

Hy·(HC+HE)-1

Vy·

(Vm

ax)-1

Fig. 2. Expected ideal sensor response as a function of applied

magnetic field strength. The dotted line shows the approxima-

tion of Eq. (6) V / Hy ðHC þ HEÞ1:

the current through the sensor, the field from otherbeads, and possibly the fringing field from thesensor itself. It will be assumed below that H isessentially in the y-direction. The magnetic fieldoutside a homogeneously magnetized sphere isidentical to the dipole field

Hdip ¼m

4pr3ð3rðm rÞ mÞ, (8)

where ^ denotes unit vectors, r ¼ rr is a vectorconnecting the centre of the bead to a pointP on the sensor, m ¼ wVbeadH is the beadmoment and Vbead is the bead volume. Fromthis, a crude estimate of the order of magnitudeof the dipole field experienced by the sensor isHwVbeadð4pz3Þ1; where z is the normaldistance between the centre of the bead and thesensor surface. The error made by this assumptionbecomes smaller when the sensor dimensionbecomes comparable to or smaller than z. Definingthe influenced area, Abead, of the sensor as thecross-section of the bead and averaging the dipolefield over the entire sensor surface, one can findthat the crude estimate given above should bemultiplied by a factor 0.38 to be correct [15]. Thus,defining the fraction of the sensor area influencedby the bead as f ¼ Abead=Asensor; the total field

ARTICLE IN PRESS


experienced by the sensor from N independentlyacting beads is

Hy Happ 0:38fNHwVbead

4pz3, (9)

where Happ is the applied external field and it isreminded that H is the total magnetic field on abead in the y-direction. Hence, the presence of thebeads reduce the effective sensitivity of the sensorto an external field by a number proportional tothe number of beads and the voltage drop over thesensor is estimated to

V y ¼ S0HyIx V 0 1H

Happ

0:38fNwVbead

4pz3

,

(10)

where V 0 ¼ S0HappIx:

3. Sensor fabrication and characterization

The PHE sensors are fabricated using con-ventional clean room fabrication methods. Thesensor layer structure of Ta(30 A)–NiFe(50 A)–MnIr(200 A)–NiFe(200 A)–Ta(30 A), presented inFig. 4, is deposited by ion beam deposition on topof a 300 silicon wafer passivated by Al2O3, and aTiWN(150 A) anti-reflecting layer is sputtered ontop. The 50 A NiFe layer is included to ensureproper growth conditions for the following MnIrlayer. For ion beam deposition conditions seeGehanno et al. [13]. Here, MnIr stands forMn76Ir24 and NiFe stands for Ni80Fe20. Duringion beam deposition a homogeneous magneticfield of 40Oe is applied in order to form an easymagnetic direction. The sensors are patterned by

NiFe (50 Å)

NiFe (200 Å)

MnIr (200 Å)

Substrate

Ta (30 Å)

Fig. 4. Sensor material layer structure. The bottom Ta(30 A)

layer is for adhesion and epitaxial growth of NiFe(50 A), which

ensures epitaxial growth of the antiferromagnetic MnIr(200 A),

which pins the sensor layer of NiFe(200 A), top Ta(30 A)

hinders corrosion.

direct laser writing into photoresist and the excessmaterial is ion-milled away in a physical dry-etch.Current leads and bonding pads of 0.3 mm thick Alare defined in final lithography, physical vapourdeposition, and lift-off steps. The whole wafer isthen passivated by 0.2 mm sputtered SiO2, and thecontact pads are opened by reactive ion etching.Finally, the wafer is diced and wire bonded to theelectrical contacts.Fig. 5 shows a micrograph of a sensor. A dotted

frame indicates the sensitive area of the sensor, butbeads from the vicinity can contribute to the signalwith reduced magnitude according to the r3

dependence of the dipole field.Fig. 6 shows vibrating sample magnetometer

data obtained on a continuous film of the stackmeasured with the applied field along the deposi-tion field. It is seen that the hysteresis loopessentially consists of two hysteretic signals, oneaccounting for E83% of the film moment withHE ¼ 41Oe and HC ¼ 18Oe and another ac-counting for E17% of the film moment withHE ¼ 274Oe and HC ¼ 21Oe: The two signals areattributed to the 200 and 50 A NiFe layers,respectively. In addition, there is a small hysteretic

Fig. 5. Top view micrograph of the planar Hall sensor. The

cross is made of magnetic layers (see layer structure in Fig. 4),

and the central area marked by a dotted frame is the

10mm 10 mm sensitive area of the sensor. Current leads are

made of 0.3mm thick Al.

ARTICLE IN PRESS

-400 -300 -200 -100 0 100 200 300 400-150

-100

-50

0

50

100

150

HC = 18 Oe

HE = 41 Oe

HC = 21 Oe

HE = 274 Oe

m (

µem

u)

H (Oe)

Fig. 6. Vibrating sample magnetometer data on the exchange-

biased permalloy film (see Fig. 4 for details on the material

structure) measured with the applied field along the easy

direction.

200

400

600

800

1000

1200

-50 -25 0 25 50

Vmax = 515 µVSlope =15.3 µV(Oe)-1 = 3.06 µV(Oe.mA)-1

Applied field (Oe)

Planar Hall Sensor (10 µm×10µm)Calibration, I = 5 mA

Sig

nal (

µV)

Fig. 7. Calibration curve for the planar Hall effect sensor. Vmax

is the highest signal minus the offset value at H ¼ 0Oe; but thecurve is shifted from H ¼ 0Oe due to a remanent magnetiza-

tion of the coil used to produce the applied field. Hence, Vmax is

measured as 12peak-to-peak signal. The slope of the linear

region gives the sensitivity of the sensor as S ¼

3:06mVðOemAÞ1 linear to within 2.8%. At H ¼ 15Oe; Vy ¼

500mV:

Table 1

Experimental and theoretical values for sensor calibration

Property Experimental Theoretical

RDC 51:9O 25OAMR (DR R1) 1.3% 1.9%

Vmax (Eq. (3) for

j ¼ p=4)515mV 458mV

HðV ¼ VmaxÞ 44Oe 47Oe

S0 3.06mV(OemA)1 3.86mV(OemA)1


impurity near zero field of unknown origin. Notethe enhancement of the coercive field with respectto that of unbiased permalloy and that the value ofHE þ HC to enter in Eq. (7) for the 200 A biasedpermalloy layer is 59Oe. This layer is expected todominate the electrical response of the sensor.The electrical characterization of the PHE

sensors was carried out by measuring the DCsensor response using a digital multimeter asfunction of the magnetic field provided by electro-magnetic coils. The sensor voltage measured forfields between 50Oe for a sensing current Ix ¼

5mA is shown in Fig. 7. It is seen that there is asignificant additional voltage offset due to, e.g.,imperfections of the lithographic process and asmall field offset due to a remanent field of theelectromagnet. Though present throughout themeasurements, the voltage offset does not influ-ence the sensitivity of the PHE sensor and is hencesubtracted from the following results. It is seenthat the general trend of the signal is in agreementwith that expected from the simple theory (Fig. 2).The low-field part of the curve (between +15 and20Oe applied field strength) is linear to within2.8% with a slope of S0 ¼ 3:06mVðOemAÞ

1: Theslope is reproducible both for the particular sensorand for all sensors on the same wafer. Table 1summarizes the experimental characteristics of thesensor and the theoretical predictions for compar-

ison. In the theoretical estimates of RDC and theAMR, based on the values for AMR of NiFe andresistivities found in [13], shunting of the currentthrough the other layers is taken into account. Forthe estimate of Vmax and the slope, only theNiFe(200 A) layer is considered; though theNiFe(50 A) might affect the signal, its contributionshould be much smaller than that from the thicklayer.

4. Detection of magnetic beads

In a recent paper [16], we have demonstratedthat the present sensors are feasible for the

ARTICLE IN PRESS

0

60

50

40

30

20

10

-10

-200 200 400 600 800 1000 1200

Time (s)

I =15 mA, Happ =-15Oe, beads: 2µm

Sig

nal (

µV)

Fig. 8. Bead detection in an applied field, Happ ¼ 15Oe and a

sensing current of Ix ¼ 15mA: A voltage offset of 1483mV is

subtracted. At times t ¼ 250 and 750 s the beads are added onto

the chip, and the signal changes according to Eq. (10) until it

reaches a saturation value. At times t ¼ 550 and 1025 s the

beads are washed off the sensor, and the signal returns to the

baseline.


detection of commercial 2 mm Micromer-M and250 nm Nanomag-D magnetic beads (Micromod,Germany [17]). These experiments were performedby measuring the voltage drop across the sensorwhile droplets containing magnetic beads wereintroduced to the sensor and subsequently washedoff. A sensing current of 10mA and an appliedexternal field of 15Oe were used. From acomparison of a direct counting of beads flowingby the sensor under an optical microscope with thecorresponding sensor signals, it was estimated thateach 2 mm Micromer-M bead contributed with asignal of 0.3 mV. The corresponding noise level ofthe unshielded DC measurements was of the samemagnitude.Here, we present experiments on the detection of

2 mm Micromer-M beads with and without apply-ing an external field from the electromagnetic coilsof 15Oe. The properties of the Micromer-Mbeads are summarized in Table 2. A beadmagnetized by 15Oe gives rise to a field justbelow the bead of +1.5Oe. The contribution tothe field outside the sensor from the sensingcurrent can be estimated using Ampere’s law to

HsenseðzÞ Ix

2w þ 2pz, (11)

where w is the width of the current line. At z ¼

1:5mm; which is a typical value for the first layer of2 mm beads, this field amounts to E0.4Oe per mAapplied sensing current.Fig. 8 shows an example of a bead detection

experiment performed in an applied field of15Oe. The current applied to the sensor is Ix ¼

15mA giving rise to an estimated field of 6Oefrom the sensing current at the first monolayer ofbeads. At time t ¼ 250 s beads are added and thesignal changes as the beads settle on top of thesensor, at time t ¼ 550 s the beads are washed off

Table 2

Physical properties of Micromer-M beads [16]

Micromer-M beads

Diameter 2 mmConcentration 425mgml1

Density (r) 1.4 g cm3

w (measured at INESC-MN) 0.370.1 [SI]

the sensor and the signal returns to its previousvalue. At times t ¼ 750 and 1025 s, respectively,the experiment is repeated yielding the same resultthough with a little higher saturation signal. Anestimate of the sensor coverage assuming that thedipole fields from the beads are independent andthat the beads are close packed on the sensorsurface yields E1.5 monolayers. The electricalnoise level of the measurement is E250 nV.Fig. 9 shows detection measurements of beads

without applying a magnetic field. The currentthrough the sensor, Ix ¼ 10mA; corresponds to afield 1.5 mm above the sensor of E4Oe. At timet ¼ 580 s a dilute bead solution is added to the chipand at time t ¼ 660 s the beads are washed off thesensor. At times t ¼ 675 and 795 s groups of beadsare observed on top of the sensitive area. The firstgroup constitutes two beads, the other four beads.They are removed by flushing with water at timest ¼ 700 and 810 s, respectively. At time t ¼ 840 s aconcentrated solution of beads is added to the chipand the signal saturates at approximately V ¼

30mV; then the beads are removed between t ¼

920 and 950 s and the signal returns to its baseline.The electrical noise level of the measurement isE250 nV.These measurements show that the mag-

netic field generated from the sensing current is

ARTICLE IN PRESS

0

10

-10

20

30

40

4 2

Sig

nal (

µV)

dilute solution

conc. solution

Time (s)600 700 800 900 1000

I =10 mA, Happ =0 Oe, beads: 2µm

Fig. 9. Bead detection in zero applied field, Happ ¼ 0Oe and a

sensing current of Ix ¼ 10mA: A voltage offset of 1775mV is

subtracted. The sensor senses a diluted bead solution, then few

beads (2 and 4), and a concentrated bead solution with a

saturation signal of V ¼ 30mV:


sufficient to yield a significant response from thebeads. Due to the inhomogeneous nature of thisfield, it will also assist in the collection of beads ontop of the sensor. The fact, that the peaks for twoand four beads are clearly distinguishable indicatesthat the sensors are feasible for the detection of afew beads or a single bead even in the absence ofan externally applied magnetizing field. It is notedthat the noise level of the unshielded DCmeasurements is expected to be improved signifi-cantly by use of lock-in technique. Although thesensor signal of PHE sensors is lower compared toequivalent spin valve and GMR sensors, the 1=f

dominated low-frequency noise level is even lowerdue to the higher number of charge carriers suchthat the signal-to-noise ratio turns out to be afactor of 3–4 higher [15]. Thus, PHE sensors arepromising candidates for the detection of singlemagnetic beads with diameters well below 1 mm.Similar experiments have been carried out for a

number of sensing currents between 5 and 15mAand an analysis of the sensor response andestimates of the number of beads collected on thesensor using Eq. (10) with the magnetizing fieldgiven by Eq. (11) yield roughly a constant cover-age of 1.4 monolayers and no clear dependence onthe applied sensing current is observed. This islikely due to the rather high initial concentration

of beads in the fluid such that the sensor isessentially saturated with beads and that theaddition of more beads does not lead to asignificant signal increase due to the r3 decay ofthe dipole field from the beads. Also, it is difficultto exactly repeat experimental conditions (fluidamounts, settle times, number of beads, etc.) fromone experiment to the next. Experiments onsystems integrated with microfluidic channels withvery dilute bead solutions and a constant fluid flowwill probably aid in the clarification of thecontribution of the sensing current to the captureefficiency and bead signal.

5. Conclusion

The planar Hall effect sensor is demonstratedcapable of detecting the presence of 2 mm Micro-mer-M beads. These beads are used for attachingbiomolecules in biological applications such asbiosensing, separation and purification protocols.The bead surfaces are covered with streptavidin toenable binding of DNA or proteins. Hence, with asuitable biochemical coating of the sensor surface,the planar Hall effect sensor can detect thepresence of DNA or proteins via the presence ofthe magnetic bead. The sensor is demonstratedsensitive enough to detect very few beads. Due tothe simple fabrication scheme, the planar Hallsensor can be easily integrated into lab-on-a-chipsystems, and the demonstration of bead detectionin the field generated by the sensing current, i.e.,without applying external fields, is promising forchip integration.Due to the cross geometry, the planar Hall effect

sensor uses its entire active surface for bead andbiomolecule detection, which is not the case formeandering-type GMR or spin valve sensors. Inaddition, the sensing geometry, compared with theordinary Hall sensor, utilizes all the magnitude ofthe magnetic dipole field induced by the bead.Another advantage, not exploited here, is that thetheoretical signal-to-noise ratio of planar Halleffect sensors is higher than those of correspond-ing GMR and spin-valve sensors [15].Future work includes demonstration of the

planar Hall effect sensor as a biosensor, and

ARTICLE IN PRESS


integration with fluidic biochips. Another prospectis to demonstrate detection of single nanometer-sized beads.

References

[1] D.R. Baselt, G.U. Lee, M. Natesan, et al., Biosensors

Bioelectron. 13 (1998) 731.

[2] R.L. Edelstein, C.R. Tamanaha, P.E. Sheehan, et al.,

Biosensors Bioelectron. 14 (2000) 805.

[3] M.M. Miller, P.E. Sheehan, R.L. Edelstein, et al., J. Magn.

Magn. Mater. 225 (2001) 138.

[4] D.L. Graham, H. Ferreira, J. Bernardo, et al., J. Appl.

Phys. 91 (2002) 7786.

[5] H.A. Ferreira, D.L. Graham, P.P. Freitas, et al., J. Appl.

Phys. 93 (2003) 7281.

[6] D.L. Graham, H.A. Ferreira, P.P. Freitas, et al., Biosen-

sors Bioelectron. 18 (2003) 483.

[7] G. Li, V. Joshi, R.L. White, et al., J. Appl. Phys. 93 (2003)

7557.

[8] P.-A. Besse, G. Boero, M. Demierre, et al., Appl. Phys.

Lett. 80 (2002) 4199.

[9] F. Montaigne, A. Schuhl, F.N. Van Dau, et al., Sensors

Actuators 81 (2000) 324.

[10] F.N. Van Dau, A. Schuhl, J.R. Childress, et al., Sensors

Actuators A 53 (1996) 256.

[11] L. Ejsing, M.F. Hansen, A.K. Menon, Planar Hall effect

magnetic sensor for microbead detection, Extended

Abstract, Eurosensors 2003, 2003, p. 1095.

[12] R.C. O’Handley, Modern Magnetic Materials, Wiley, New

York, 2000.

[13] V. Gehanno, P.P. Freitas, A. Veloso, et al., IEEE Trans.

Magn. 35 (1999) 4361.

[14] A. Schuhl, F.N. Van Dau, J.R. Childress, Appl. Phys. Lett.

66 (1995) 2751.

[15] P.P. Freitas, H.A. Ferreira, D.L. Graham, et al., in: M.

Johnson (Ed.), Magnetoresistive Biochips, Elsevier, Am-

sterdam, 2004.

[16] L. Ejsing, M.F. Hansen, A.K. Menon, et al., Appl. Phys.

Lett. 84 (2004) 4729.

[17] http://www.micromod.de.

http://www.micromod.de

Biosensors and Bioelectronics 21 (2006) 1327–1332

Short communication

Immobilisation of DNA to polymerised SU-8 photoresist

Rodolphe Mariea, Silvan Schmidb, Alicia Johanssona, Louise Ejsinga, Maria Nordstroma,Daniel Hafligera, Claus BV Christensena, Anja Boisena, Martin Dufvaa,∗

a Department of Micro- and Nano-Technology, Technical University of Denmark, Bldg. 345, DK-2800 Lyngby, Denmarkb Department of Mechanical and Process Engineering, Nanotechnology Group, Swiss Federal Institute of Technology,

ETH Zentrum CLA, CH-8092 Zurich, Switzerland

Received 12 January 2005; received in revised form 10 March 2005; accepted 11 March 2005Available online 31 March 2005

Abstract

SU-8 is an epoxy-based photosensitive resist, which is currently used for a large variety of MEMS and lab-on-a-chip applications. Here,we demonstrate a one-step process to functionalise SU-8 with DNA probes. The immobilisation procedure relies on direct coupling of DNAto SU-8 and resulted in surfaces with functional capture probe densities of approximately 10 fmol/mm2 as determined by hybridisation assaysw hyde coatedg ac solutiond©

K

1

nowsmoDbta(w

b

nning,,;et

ed,ebleitsching

he-re-

0d

ith fluorescent labelled target molecules. A comparable density of functional capture probes was measured on commercial aldelass. DNA probes did not decrease in hybridisation performance after 10 min incubation in water at 98C prior to hybridisation, indicatingovalent bond between DNA and SU-8. Finally, DNA microarrays of high quality were obtained on SU-8 by contact printing of probeirectly on SU-8 demonstrating a simple method for the implementation of microarrays in microsystems.2005 Elsevier B.V. All rights reserved.

eywords: SU-8; DNA; Immobilisation; Photoresist; Hybridisation; Microstructure; MEMS; Lab-on-a-chip

. Introduction

A future goal of the lab-on-a-chip is to integrate every stepeeded for detecting an analyte in the sample. The analysisf a sample is often based on size separations or interactionith a probe molecule either in solution or attached to a solidupport. Using high throughput solid phase reactions likeicroarray, tens of thousands of analytes can be detected inne batch process. An efficient procedure for immobilisingNA or proteins to a solid support is essential for solid phaseiochemical reactions. DNA has been successfully attached

o glass (Zammatteo et al., 2000), oxidized silicon (Chrisey etl., 1996), silicon wafers (Strother et al., 2000), gold surfacesSteel et al., 2000) and PMMA (Fixe et al., 2004a, 2004b)hich all are relevant surfaces that can be micro-fabricated.Because of its attractive mechanical properties, the epoxy-

ased photoresist SU-8 has become widely used for the fab-

∗ Corresponding author. Tel.: +45 45 25 57 00; fax: +45 45 25 77 62.E-mail address: [email protected] (M. Dufva).

rication of mechanical structures such as probes for scaprobe microscopy (Genolet et al., 1999, 2001). FurthermoreSU-8 can be fabricated into optical waveguides (Lee et al.2002) and into microfluidic structures (Chuang et al., 2003Heuschkel et al., 1998, Jackman et al., 2001; Seidemannal., 2002a, 2002b). Those functionalities can be combinand integrated to build complete biosensors (Calleja et al.2003; Mogensen et al., 2003). Since SU-8 is an attractivmaterial for MEMS and micro fluidic devices, it is desirato develop procedures for immobilising DNA directly onsurface. Here we describe a one-step procedure for attaDNA directly onto cured micropatterned SU-8.

2. Materials and methods

2.1. Processing of photoresist

Commercially available SU-8 5 (glycidyl ether of bispnol A) (Microresist Technology, Berlin, Germany) and

956-5663/$ – see front matter © 2005 Elsevier B.V. All rights reserved.oi:10.1016/j.bios.2005.03.004

1328 R. Marie et al. / Biosensors and Bioelectronics 21 (2006) 1327–1332

Table 1Probes and target molecules

Probe name Sequence 5′-Endmodification

MT(19) 5′-CCACGCTCATTGACAAGCT-3′ NoneWT(19) 5′-CCACGCTCATCGACAAGCT-3′ NoneMT(19)-NH2 5′-CCACGCTCATTGACAAGCT-3′ AmineWT(19)-NH2 5′-CCACGCTCATCGACAAGCT-3′ AmineWT(60) 5′-AAT TGC GAC AGT TGA AGA

TGC TGC CAC GCT CAT CGA CAAGCT TGT CTC TGC GAA GTCATC-3′

None

WT(60)-NH2 5′-AAT TGC GAC AGT TGA AGATGC TGC CAC GCT CAT CGA CAAGCT TGT CTC TGC GAA GTCATC-3′

Amine

W54-Cy3 5′-GAT GAC TTC GCA GAG ACAAGC TTG TCG ATG AGC GTGGCA GCA TCT TCA ACT GTCGCA ATT-3′

Cy3

“WT” refers to wild type probes and “MT” refers to single mismatch mutantprobes.

sist thinner (gamma butyrolactone) were mixed in a 2:1 ra-tio. An SU-8 thickness of 0.8m was obtained when spin-coating the resist onto a glass wafer at a spinning speed of3000 rpm. After spinning, the excess solvent was evaporatedin a soft bake for 3 min at 65C and 2 min at 95C. TheSU-8 was exposed to UV-light through a mask for 30 s at9 mW/cm2 and baked for 1 h at 90C. The wafers were de-veloped 2 min in 1-methoxy-2-propyl acetate (PGMEA) andrinsed in isopropanol. Alternatively, SU-8 2002 (with the sol-vent cyclopentanone) was used to pattern SU-8 as 20m widelines on glass. An SU-8 thickness of 1.7m was obtainedby spin-coating the SU-8 2002 at 5000 rpm. The excess sol-vent was evaporated in a soft bake for 1 min at 60C and1 min at 90C. The SU-8 was exposed to UV-light througha mask for 50 s at 8.1 mW/cm2 and baked for 5 min at 60Cand 10 min at 90C. The wafers were developed 2 min inPGMEA and rinsed in isopropanol. Free hanging SU-8 can-tilevers were fabricated as described previously (Calleja et al.,2004).

2.2. Immobilisation of DNA

DNA probes (Table 1) were diluted to 10M in 1 M potas-sium phosphate pH 7.4 (Sigma–Aldrich Co., Steinheim, Ger-many) and manually spotted at a volume of 0.5L per spoto As-s olu-t etix,N ng-i rs-i bes tmo-sw ogens

2.3. Hybridisation

The SU-8 samples (SU-8, structured SU-8 and cantilevers)were hybridised for 1 h at 37C with a solution containing10 nM W54-Cy3 (Table 1), 5× sodium chloride sodium cit-rate (SSC) (Promega Corporation, Madison, US) and 0.1%sodium dodecylsulfate (SDS) (Sigma–Aldrich). The SU-8 samples were washed either in low stringent conditions(10 min in 1× SSC containing 0.1% SDS at room tempera-ture) or at high stringency (10 min in 0.1× SSC containing0.5% SDS at 45C). The arrays were rinsed in a buffer with-out SDS and dried with a gentle stream of nitrogen. Aldehydecoated slides were hybridised and washed using the same pro-cedure.

2.4. Fluorescence detection

The arrays were scanned using a CCD scanner (Array-WorX, Applied Precision, Issaquah, WA, USA). Mean valuesof the fluorescence density were obtained using the analy-sis software Data Inspector (Arrayworx). Spot intensity val-ues on microarrays were obtained using the ScanAnalyzesoftware (Eisen and Brown, 1999). The quantification ofthe density of the hybridisations were obtained by report-ing the mean values of fluorescence density on a calibra-tion curve made by spotting a dilution series of Cy3 labelledo and( teds

3

3

ick-n e op-t ssay.A %h while1 nd6 thatt ordert renti U-8t tio(

3

nd-i oupsw sur-f edt n-c

n polymerised SU-8 or on aldehyde coated slides (CELociates Inc., Pearland, TX, USA). Alternatively the sions were micro-spotted using a pin arrayer (Qarray, Genew Milton, Hampshire, UK) at 50% humidity. Free ha

ng SU-8 cantilevers were functionalised by totally immeng the cantilevers in a few microliters of the DNA proolution. The samples were incubated in a humid aphere for 1 h at 37C and washed for 10 min in MilliQater. The arrays and structures were dried in a nitrtream.

ligonucleotide (W54-Cy3) on an SU-8 substrate by h0.5L/spot) or by microarray printing on aldehyde coalides (1 nL/spot).

. Results and discussion

.1. Auto-fluorescence of SU-8

SU-8 was deposited onto glass wafers at different thesses and scanned in the Cy3 channel to evaluate th

ical properties of SU-8 in a Cy3-based fluorescent an SU-8 layer of 0.8m thickness had approximately 20igher fluorescence signal than the glass wafer alone,.7, 12 and 120m thick layers of SU-8 had 50, 400 a600% more signal than glass, respectively, indicating

he thinnest SU-8-layer as possible should be used ino minimise the background signal. This effect was appan hybridisation reaction in which the thinnest layer of Sested (0.8m) resulted in the highest signal to noise radata not shown).

.2. Immobilisation of DNA to SU-8

Investigating SU-8, we suspected the ability of DNA bing by condensation of primary and secondary amine grith the non-cross linked epoxy rings present on the

ace. Determination of optimal binding conditions showhat DNA could efficiently be immobilised using a “wet” iubation in which the DNA was spotted onto SU-8 in 0.5L

R. Marie et al. / Biosensors and Bioelectronics 21 (2006) 1327–1332 1329

Fig. 1. Immobilisation of DNA to SU-8. (A) Heat stability of aminated 60 bpprobe immobilised on SU-8. Three samples were washed at 25, 50 and 98C,respectively, before being hybridised to W54-Cy3. Note, the fluorescentsignal was collected for 10-fold shorter time inFig. 1A than inFig. 1C andFig. 2hence, the lower fluorescent signal. (B) Image of spots on SU-8 andaldehyde coated slides after hybridisation reactions. W60-NH2 probe wasprinted by contact printing according to material and methods section. (C)The quantified fluorescence hybridisation signal from DNA-capture probesdispensed on SU-8 and aldehyde coated slides using contact printing. “WT”and “MT” refers to wild type and mutant 19′mer probes, respectively.

drops followed by incubation in a humid environment for 1 hat 37C. A baking or UV-cross-link procedure did not im-prove DNA binding to the SU-8 surface (data not shown).To estimate the amount of probes that were available for hy-bridisation, the SU-8 pieces with immobilised DNA were hy-bridised with Cy3 labelled complementary target molecules(W54-Cy3,Table 1). The hybridised density was determinedto be 6 fmol/mm2 and no cross-hybridisation was observed tonon-related probes (see also results from microarrays printedon SU-8,Fig. 1C). The DNA molecules were likely to becovalently attached to the SU-8 surface since 10 min incu-bation at 98C prior to hybridisation resulted in no loss offunctional probes (Fig. 1A).

3.3. DNA microarrays on SU-8

Fabrication of DNA microarray on SU-8 was investigatedby spotting 60 and 19 bp long probes, with and withoutamino-modification, onto SU-8 and aldehyde coated slides.The hybridised density of the 60 bp probes immobilised onSU-8 was between 8 and 13 fmol/mm2 while microarrays oncommercial aldehyde coated glass had a hybridised density of9 fmol/mm2. These hybridisation densities were comparablewith those obtained previously on 2D microarray substrates(Dufva et al., 2004; Fixe et al., 2004a, 2004b). The microar-ray spots on SU-8 were perfectly circular with a diameterof 100m and appeared more homogeneous than the cor-responding spots on the aldehyde coated slides, which haddiameters of 150m (Fig. 1B). The smaller spot size on SU-8 is most likely due to the higher hydrophobicity of SU-8 ascompared to aldehyde coated slides. These results demon-strate that high quality microarrays can be fabricated usingSU-8 as substrate, which enables easy incorporation of mi-croarrays into microfluidic systems.

3.4. Specificity of immobilised probes

The specificity of immobilised probes was investigatedby hybridising W54-Cy3 to 19 base pair probes differingin only one base in the middle. After stringency wash, W54-C obes( 2)i et-t isedo atchp atesr theW dp ne oft heser ifiedp UVc tion(

3b

beh besi lides( re-s di-fi ltm havem thus,h NAc

at-t not

y3 target bound three-fold better to the perfect match prWT(19)-NH2) than to the mismatch probes (MT(19)-NHmmobilised on SU-8 while W54-Cy3 bound four-fold ber to the perfect match amino-modified probes immobiln aldehyde slides compared to the corresponding mismrobe (Fig. 1C). The unmodified probes on both substresulted in ratio of four between the signal obtained from

T probe and the MT probe (Fig. 1C). Thus, also unmodifierobes that must be attached to SU-8 through at least o

he bases, can discriminate a mismatch hybridisation. Tesults corroborate previous findings that short unmodrobes, in this case immobilised to an agarose film byross-linking, could discriminate a mismatch hybridisaDufva et al., 2004).

.5. Influence of amino modification and nature of theond

An amino modification in the terminal of the oligo proad no influence on the hybridised signal from 60 bp pro

mmobilised on SU-8 substrates or the aldehyde coated sFig. 1C). By contrast, the amino modified 19 bp probesulted in a four to five-fold higher signal than the unmoed 19 bp probes (Fig. 1C). This apparently conflicting resuight be explained by the facts that the 60 bp probesore bases that can form bonds with the substrate andave a larger likelihood to be attached through intra-Dhain bonds than 19 bp probes.

The above results strongly indicate that the DNA isached to the SU-8 by a covalent bond. However, it is


clear which functional groups are involved in the bond be-tween DNA and SU-8. It is not likely that the DNA is con-nected to free epoxy groups since pre-treatment of SU-8 withepoxy ring opening agents (ethanolamine, HCl or acetic acid)did not influence probe binding and functionality (data notshown). A chrome etch (Cr etch 1020 (Transene Co Inc.))did, however, block the SU-8 probe binding as assessedby hybridisation reaction indicating that a specific chemicalgroup in the SU-8 is involved. Although the bond betweenDNA and SU-8 is not well defined, the bond could with-stand boiling temperatures for 10 min, which is importantfor de-hybridisation reactions to regenerate sensor surfaces.

It should be noted that unmodified DNA bound strongly toSU-8 indicating that DNA assays performed in SU-8 struc-tures could suffer as a result of SU-8 sequestering sampleDNA. Our results indicate that SU-8 surfaces were pacifiedby the hybridisation solution which was made of 5× SSCand 0.1% SDS. We speculate that SDS binds to SU-8 andthereby blocks unspecific binding of target DNA to SU-8.

3.6. Immobilisation of DNA to structured SU-8

It would be desirable to be able to direct the DNA to cer-tain parts of a microfluidic system, for example at sites of

Fscq

ig. 2. Immobilisation of DNA to structured SU-8. (A) Fluorescence image scanned at the maximum resolution of the scanner (5m2/pixel). (B) Image of hapture probe solution (upper left panel) and cantilevers exposed to DNA cauantification of the fluorescent signal on the cantilevers.

howing the 20m wide lines of SU-8 (bright) on glass (dark). The image wasybridisation signals from SU-8 cantilevers that were not exposed to DNA-pture probe solution (upper right panel) prior to hybridisation. Lower panel shows

R. Marie et al. / Biosensors and Bioelectronics 21 (2006) 1327–1332 1331

detection. To demonstrate the possibility of such localisa-tion, SU-8 was structured as 20m wide lines on glass with20m spacing. WT(60)-NH2 was immobilised on the struc-tured SU-8-sample and hybridised with W54-Cy3. Most ofthe hybridisation signals were localised on the SU-8 lines(Fig. 2A). The hybridised density on the SU-8 lines was es-timated to be 9 fmol/mm2 indicating good reproducibility ofthe immobilisation method, especially considering that theSU-8 samples were made of a SU-8 2002 instead of SU-85 as inFig. 1. Since SU-8 5 and SU-8 2002 are based ontwo different solvents it is likely that it is chemical groupsbelonging to the SU-8 polymer chain that are involved inbinding DNA. There was up to 20-fold difference in sig-nals on the SU-8 lines as compared to the area betweenthe lines. However, there was usually >100-fold differencein functional probes on SU-8 as compared to unmodifiedglass, indicating the presence of residual SU-8 on the glassafter structuring. However, AZ5214E resist (Hoechst, Ger-many) binds DNA approximately as an unmodified glasswafer (data not shown) suggesting the possibility of fabricat-ing localised SU-8 structures with high selectivity of probebinding using another layer of AZ5214E resist on top ofthe SU-8.

3.7. Immobilisation of DNA to SU-8 cantilevers

vantM ion-a ers( -a andhl tot onc isa-t nc-t elyt f thec olu-t on-s gn is-a weti

in-fl hang-i otheree im-me . Ag ef-f en-s sula-t esss

4. Conclusions

SU-8 could be functionalised with DNA capture probeswith functional probe densities of 6–13 fmol/mm2. The im-mobilised DNA has excellent specificity down to single basepair level indicating that probes immobilised on SU-8 couldbe used for genotyping. The stable bond between SU-8 andDNA makes regeneration of sensor surfaces possible. The im-mobilising of DNA directly to SU-8 avoids several processsteps otherwise needed to link DNA to glass and simplifiesintegration of DNA in biosensors like cantilevers and mag-netoresistive biochips.

Acknowledgements

We thank Dorthe Thybo Ganzhorn for spending time in thecleanroom fabricating the photoresist substrate. This projectwas funded by the Danish research council (Grant #2014-00-0003, DABIC and Grant #26-02-0280, Polymeric can-tilevers)

References

Calleja, M., Nordstrom, M., Alvarez, M., Tamayo, J., Lechuga, L., Boisen,s for

C mericstem.tical

C a, L.,ppli-

C t ofRes.

C ovelhick-

D nsen,ighly

E pres-

E itas,bead

F onal-NA

F 04b.nto195.

G ., defor

G j, N.,inedRev.

To demonstrate that DNA can be attached to releEMS structures, free hanging cantilevers were functlised with amine modified DNA probes. The cantilev20m wide, 200m long and 5m thick) were functionlized by dipping them in capture probe DNA solutionybridised with a complementary 60′-mer long target DNA

abelled with Cy3′. A reference chip was only exposedarget DNA solution. Hybridisation was only observedantilevers exposed to probe solution prior to hybridion (Fig. 2B) indicating that SU-8 cantilevers can be fuionalised with DNA capture probes. It is, however, likhat the cantilevers are functionalised on both sides oantilever but this can be avoided by spotting probe sion onto the chip using a dispensing machine. As demtrated above (Fig. 1) the DNA immobilised by spottinL droplets onto SU-8 functions equally well in hybridtion reactions as compared to DNA immobilised by

ncubation.The drawback of SU-8 being auto-fluorescent will not

uence sensors based on electrical readouts like freeng SU-8 cantilevers and magnetoresistive biochips, anxample of a biosensor that we currently pursue (Ejsingt al., 2004). SU-8 based devices can be modified byobilising thiolated DNA on gold coated SU-8 (Callejat al., in press). This, however, has several drawbacksold layer can for example cause unwanted bimorph

ects and short-circuiting of electrical wires. On both sors types, SU-8 would be used as a combined inion/immobilisation layer, hereby saving several procteps.

A., in press. Highly sensitive polymer-based cantilever-sensorDNA detection. Ultramicroscopy.

alleja, M., Rasmussen, P., Johansson, A., Boisen, A., 2003. Polymechanical sensors with integrated readout in a microfluidic syIn: Proceedings of the SPIE—the International Society for OpEngineering 5116, pp. 314–321.

alleja, M., Tamayo, J., Johansson, A., Rasmussen, P., LechugBoisen, A., 2004. Polymeric cantilever arrays for biosensing acations. Sens. Lett. 1, 20–24.

hrisey, L.A., Lee, G.U., O’Ferrall, C.E., 1996. Covalent attachmensynthetic DNA to self-assembled monolayer films. Nucleic Acids24, 3031–3039.

huang, Y.-J., Tseng, F.-G., Cheng, J.-H., Lin, W.-K., 2003. A nfabrication method of embedded micro-channels by using SU-8 tfilm photoresists. Sens. Actuators A: Phys. A 102, 64–69.

ufva, M., Petronis, S., Bjerremann Jensen, L., Krag, C., ChristeC., 2004. Characterization of an inexpensive, non-toxic and hsensitive microarray substrate. Biotechniques 37, 286–296.

isen, M.B., Brown, P.O., 1999. DNA arrays for analysis of gene exsion. Methods Enzymol. 303, 179–205.

jsing, L., Hansen, M., Menon, A., Ferreira, H., Graham, D., FreP., 2004. Planar Hall effect sensor for magnetic micro- and nanodetection. Appl. Phys. Lett. 84, 4729.

ixe, F., Dufva, M., Telleman, P., Christensen, C.B., 2004a. Functiization of poly(methyl methacrylate) (PMMA) as a substrate for Dmicroarrays. Nucleic Acids Res. 32, e9.

ixe, F., Dufva, M., Telleman, P., Christensen, C.B., 20One-step immobilization of aminated and thiolated DNA opoly(methylmethacrylate) (PMMA) substrates. Lab. Chip. 4, 191–

enolet, G., Brugger, J., Despont, M., Dreschler, U., Vettiger, PRooj, N., Anselmetti, D., 1999. Soft, entirely photoplastic probesscanning force microscopy. Rev. Sci. Instrum. 70, 2398.

enolet, G., Despont, M., Vettiger, P., Staufer, U., Noell, W., de RooCueni, T., Bernal, M.-P., Marquis-Weible, F., 2001. Micromachphotoplastic probe for scanning near-field optical microscopy.Sci. Instrum. 72, 3872.


Heuschkel, M.O., Buisson, B., Renaud, P., 1998. Buried microchannelsin photopolymer for delivering of solutions to neurons in a network.Sens. Actuators B: Chem. 48, 356–361.

Jackman, R.J., Floyd, T.M., Ghodssi, R., Schmidt, M.A., Jensen, K.F.,2001. Microfluidic systems with on-line UV detection fabricated inphotodefinable epoxy. J. Micromech. Microeng. 11, 263–269.

Lee, G.-B., Lin, C.-H., Chang, G.-L., 2002. Multi-cell-line micro flow cy-tometers with buried SU-8/SOG optical waveguides. In: Proceedingsof the 15th IEEE International Conference on Micro Electro Mechan-ical Systems. In Technical Digest. MEMS 2002 IEEE InternationalConference, p. 503.

Mogensen, K.B., El-Ali, J., Wolff, A., Kutter, J.P., 2003. Integration ofpolymer waveguides for optical detection in microfabricated chemicalanalysis systems. Appl. Opt. 42, 4072–4079.

Seidemann, V., Butefisch, S., Buttgenbach, S., 2002a. Fabrication andinvestigation of in-plane compliant SU8 structures for MEMS and

their application to micro valves and micro grippers. Sens. ActuatorsA: Phys. 97–98, 457–461.

Seidemann, V., Rabe, J., Feldmann, M., Buttgenbach, S., 2002b. SU8-micromechanical structures with in situ fabricated movable parts. Mi-crosyst. Technol. 8, 348–350.

Steel, A.B., Levicky, R.L., Herne, T.M., Tarlov, M.J., 2000. Immobiliza-tion of nucleic acids at solid surfaces: effect of oligonucleotide lengthon layer assembly. Biophys J. 79, 975–981.

Strother, T., Hamers, R.J., Smith, L.M., 2000. Covalent attachment ofoligodeoxyribonucleotides to amine-modified Si (0 0 1) surfaces. Nu-cleic Acids Res. 28, 3535–3541.

Zammatteo, N., Jeanmart, L., Hamels, S., Courtois, S., Louette, P., Hevesi,L., Remacle, J., 2000. Comparison between different strategies of co-valent attachment of DNA to glass surfaces to build DNA microarrays.Anal. Biochem. 280, 143–150.

planar hall sensor for influenza immunoassay · s=nc = 2 is found for the magnetic immunoassay, and...

Documents